From 576a10797b607ee9e4068218daf367b481564120 Mon Sep 17 00:00:00 2001 From: Yang Wang Date: Tue, 3 Mar 2026 21:10:11 -0500 Subject: drm/amd/pm: add missing od setting PP_OD_FEATURE_ZERO_FAN_BIT for smu v13 add missing od setting PP_OD_FEATURE_ZERO_FAN_BIT for smu v13.0.0/13.0.7 Fixes: cfffd980bf21 ("drm/amd/pm: add zero RPM OD setting support for SMU13") Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/5018 Signed-off-by: Yang Wang Acked-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c | 3 ++- drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c | 3 ++- 2 files changed, 4 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c index b32c053950c9..a8d63d4d1f6e 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c @@ -2222,7 +2222,8 @@ static int smu_v13_0_0_restore_user_od_settings(struct smu_context *smu) user_od_table->OverDriveTable.FeatureCtrlMask = BIT(PP_OD_FEATURE_GFXCLK_BIT) | BIT(PP_OD_FEATURE_UCLK_BIT) | BIT(PP_OD_FEATURE_GFX_VF_CURVE_BIT) | - BIT(PP_OD_FEATURE_FAN_CURVE_BIT); + BIT(PP_OD_FEATURE_FAN_CURVE_BIT) | + BIT(PP_OD_FEATURE_ZERO_FAN_BIT); res = smu_v13_0_0_upload_overdrive_table(smu, user_od_table); user_od_table->OverDriveTable.FeatureCtrlMask = 0; if (res == 0) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c index f08cfa510a8a..5500a0f12f0e 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c @@ -2224,7 +2224,8 @@ static int smu_v13_0_7_restore_user_od_settings(struct smu_context *smu) user_od_table->OverDriveTable.FeatureCtrlMask = BIT(PP_OD_FEATURE_GFXCLK_BIT) | BIT(PP_OD_FEATURE_UCLK_BIT) | BIT(PP_OD_FEATURE_GFX_VF_CURVE_BIT) | - BIT(PP_OD_FEATURE_FAN_CURVE_BIT); + BIT(PP_OD_FEATURE_FAN_CURVE_BIT) | + BIT(PP_OD_FEATURE_ZERO_FAN_BIT); res = smu_v13_0_7_upload_overdrive_table(smu, user_od_table); user_od_table->OverDriveTable.FeatureCtrlMask = 0; if (res == 0) -- cgit v1.2.3 From b7ac77468cda92eecae560b05f62f997a12fe2f2 Mon Sep 17 00:00:00 2001 From: Mario Limonciello Date: Wed, 4 Mar 2026 14:07:40 -0600 Subject: drm/amd: Fix NULL pointer dereference in device cleanup When GPU initialization fails due to an unsupported HW block IP blocks may have a NULL version pointer. During cleanup in amdgpu_device_fini_hw, the code calls amdgpu_device_set_pg_state and amdgpu_device_set_cg_state which iterate over all IP blocks and access adev->ip_blocks[i].version without NULL checks, leading to a kernel NULL pointer dereference. Add NULL checks for adev->ip_blocks[i].version in both amdgpu_device_set_cg_state and amdgpu_device_set_pg_state to prevent dereferencing NULL pointers during GPU teardown when initialization has failed. Fixes: 39fc2bc4da00 ("drm/amdgpu: Protect GPU register accesses in powergated state in some paths") Reviewed-by: Alex Deucher Signed-off-by: Mario Limonciello Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 6f6973e8cd53..f7309705788d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2550,6 +2550,8 @@ int amdgpu_device_set_cg_state(struct amdgpu_device *adev, i = state == AMD_CG_STATE_GATE ? j : adev->num_ip_blocks - j - 1; if (!adev->ip_blocks[i].status.late_initialized) continue; + if (!adev->ip_blocks[i].version) + continue; /* skip CG for GFX, SDMA on S0ix */ if (adev->in_s0ix && (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GFX || @@ -2589,6 +2591,8 @@ int amdgpu_device_set_pg_state(struct amdgpu_device *adev, i = state == AMD_PG_STATE_GATE ? j : adev->num_ip_blocks - j - 1; if (!adev->ip_blocks[i].status.late_initialized) continue; + if (!adev->ip_blocks[i].version) + continue; /* skip PG for GFX, SDMA on S0ix */ if (adev->in_s0ix && (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GFX || -- cgit v1.2.3 From f0e46fd06c3f7590b0f06aea3c877f441c2cbccc Mon Sep 17 00:00:00 2001 From: Sunil Khatri Date: Tue, 3 Mar 2026 22:25:57 +0530 Subject: drm/amdgpu/userq: remove queue from doorbell xarray MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In case of failure in xa_alloc, remove the queue during clean up from the userq_doorbell_xa. Signed-off-by: Sunil Khatri Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c index 7c450350847d..2716811756d2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c @@ -865,6 +865,7 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args) if (r) { drm_file_err(uq_mgr->file, "Failed to allocate a queue id\n"); amdgpu_userq_fence_driver_free(queue); + xa_erase_irq(&adev->userq_doorbell_xa, index); uq_funcs->mqd_destroy(queue); kfree(queue); r = -ENOMEM; -- cgit v1.2.3 From a978ed3d6454e4aa1a2ac74051ffd77b7d263e44 Mon Sep 17 00:00:00 2001 From: Sunil Khatri Date: Tue, 3 Mar 2026 22:30:04 +0530 Subject: drm/amdgpu/userq: remove queue from doorbell xa during clean up MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit If function amdgpu_userq_map_helper fails we do need to clean up and remove the queue from the userq_doorbell_xa. Signed-off-by: Sunil Khatri Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c index 2716811756d2..3ae7d1f5d115 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c @@ -885,6 +885,7 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args) r = amdgpu_userq_map_helper(queue); if (r) { drm_file_err(uq_mgr->file, "Failed to map Queue\n"); + xa_erase_irq(&adev->userq_doorbell_xa, index); xa_erase(&uq_mgr->userq_xa, qid); amdgpu_userq_fence_driver_free(queue); uq_funcs->mqd_destroy(queue); -- cgit v1.2.3 From 1b5cf07d80bb16d1593579ccdb23f08ea4262c14 Mon Sep 17 00:00:00 2001 From: Yang Wang Date: Tue, 3 Mar 2026 21:14:10 -0500 Subject: drm/amd/pm: add missing od setting PP_OD_FEATURE_ZERO_FAN_BIT for smu v14 add missing od setting PP_OD_FEATURE_ZERO_FAN_BIT for smu v14.0.2/14.0.3 Fixes: 9710b84e2a6a ("drm/amd/pm: add overdrive support on smu v14.0.2/3") Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/5018 Signed-off-by: Yang Wang Acked-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c index 9994d4369da8..73762d9b5969 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c @@ -2311,7 +2311,8 @@ static int smu_v14_0_2_restore_user_od_settings(struct smu_context *smu) user_od_table->OverDriveTable.FeatureCtrlMask = BIT(PP_OD_FEATURE_GFXCLK_BIT) | BIT(PP_OD_FEATURE_UCLK_BIT) | BIT(PP_OD_FEATURE_GFX_VF_CURVE_BIT) | - BIT(PP_OD_FEATURE_FAN_CURVE_BIT); + BIT(PP_OD_FEATURE_FAN_CURVE_BIT) | + BIT(PP_OD_FEATURE_ZERO_FAN_BIT); res = smu_v14_0_2_upload_overdrive_table(smu, user_od_table); user_od_table->OverDriveTable.FeatureCtrlMask = 0; if (res == 0) -- cgit v1.2.3 From 1837c76b780a4201e3d9f718e17c09b536df700f Mon Sep 17 00:00:00 2001 From: Thomas Zimmermann Date: Wed, 4 Mar 2026 13:58:38 +0100 Subject: drm/amdgpu: Move test for fbdev GEM object into generic helper Provide a generic helper that tests if fbdev emulation is backed by a specific GEM object. Not all drivers use client buffers (yet), hence also test against the first GEM object in the fbdev framebuffer. Convert amdgpu. The helper will also be useful for radeon. Reviewed-by: Alex Deucher Signed-off-by: Thomas Zimmermann Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_display.c | 21 +++----------------- drivers/gpu/drm/drm_fb_helper.c | 30 +++++++++++++++++++++++++++++ include/drm/drm_fb_helper.h | 8 ++++++++ 3 files changed, 41 insertions(+), 18 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index bef9dce2e7ea..f5cd68542442 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -1738,21 +1738,6 @@ bool amdgpu_crtc_get_scanout_position(struct drm_crtc *crtc, stime, etime, mode); } -static bool -amdgpu_display_robj_is_fb(struct amdgpu_device *adev, struct amdgpu_bo *robj) -{ - struct drm_device *dev = adev_to_drm(adev); - struct drm_fb_helper *fb_helper = dev->fb_helper; - - if (!fb_helper || !fb_helper->buffer) - return false; - - if (gem_to_amdgpu_bo(fb_helper->buffer->gem) != robj) - return false; - - return true; -} - int amdgpu_display_suspend_helper(struct amdgpu_device *adev) { struct drm_device *dev = adev_to_drm(adev); @@ -1775,7 +1760,6 @@ int amdgpu_display_suspend_helper(struct amdgpu_device *adev) list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) { struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); struct drm_framebuffer *fb = crtc->primary->fb; - struct amdgpu_bo *robj; if (amdgpu_crtc->cursor_bo && !adev->enable_virtual_display) { struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo); @@ -1790,8 +1774,9 @@ int amdgpu_display_suspend_helper(struct amdgpu_device *adev) if (!fb || !fb->obj[0]) continue; - robj = gem_to_amdgpu_bo(fb->obj[0]); - if (!amdgpu_display_robj_is_fb(adev, robj)) { + if (!drm_fb_helper_gem_is_fb(dev->fb_helper, fb->obj[0])) { + struct amdgpu_bo *robj = gem_to_amdgpu_bo(fb->obj[0]); + r = amdgpu_bo_reserve(robj, true); if (r == 0) { amdgpu_bo_unpin(robj); diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c index 05803169bed5..a883534e19e4 100644 --- a/drivers/gpu/drm/drm_fb_helper.c +++ b/drivers/gpu/drm/drm_fb_helper.c @@ -37,6 +37,7 @@ #include #include #include +#include #include #include #include @@ -1786,3 +1787,32 @@ int drm_fb_helper_hotplug_event(struct drm_fb_helper *fb_helper) return 0; } EXPORT_SYMBOL(drm_fb_helper_hotplug_event); + +/** + * drm_fb_helper_gem_is_fb - Tests if GEM object is framebuffer + * @fb_helper: fb_helper instance, can be NULL + * @obj: The GEM object to test, can be NULL + * + * Call drm_fb_helper_gem_is_fb to test is a DRM device's fbdev emulation + * uses the specified GEM object for its framebuffer. The result is always + * false if either poiner is NULL. + * + * Returns: + * True if fbdev emulation uses the provided GEM object, or false otherwise. + */ +bool drm_fb_helper_gem_is_fb(const struct drm_fb_helper *fb_helper, + const struct drm_gem_object *obj) +{ + const struct drm_gem_object *gem = NULL; + + if (!fb_helper || !obj) + return false; + if (fb_helper->buffer && fb_helper->buffer->gem) + gem = fb_helper->buffer->gem; + else if (fb_helper->fb) + gem = drm_gem_fb_get_obj(fb_helper->fb, 0); + + return gem == obj; +} +EXPORT_SYMBOL_GPL(drm_fb_helper_gem_is_fb); + diff --git a/include/drm/drm_fb_helper.h b/include/drm/drm_fb_helper.h index 05cca77b7249..ca214b5c0283 100644 --- a/include/drm/drm_fb_helper.h +++ b/include/drm/drm_fb_helper.h @@ -271,7 +271,15 @@ int drm_fb_helper_ioctl(struct fb_info *info, unsigned int cmd, int drm_fb_helper_hotplug_event(struct drm_fb_helper *fb_helper); int drm_fb_helper_initial_config(struct drm_fb_helper *fb_helper); +bool drm_fb_helper_gem_is_fb(const struct drm_fb_helper *fb_helper, + const struct drm_gem_object *obj); #else +static inline bool drm_fb_helper_gem_is_fb(const struct drm_fb_helper *fb_helper, + const struct drm_gem_object *obj) +{ + return false; +} + static inline void drm_fb_helper_prepare(struct drm_device *dev, struct drm_fb_helper *helper, unsigned int preferred_bpp, -- cgit v1.2.3 From 7ba5286aad4e670b64053d64c04dfdf575594692 Mon Sep 17 00:00:00 2001 From: Thomas Zimmermann Date: Wed, 4 Mar 2026 13:58:39 +0100 Subject: drm/radeon: Test for fbdev GEM object with generic helper Replace radeon's test for the fbdev GEM object with a call to the generic helper. Reviewed-by: Alex Deucher Signed-off-by: Thomas Zimmermann Signed-off-by: Alex Deucher --- drivers/gpu/drm/radeon/radeon_device.c | 7 ++++--- drivers/gpu/drm/radeon/radeon_fbdev.c | 17 ----------------- drivers/gpu/drm/radeon/radeon_mode.h | 5 ----- 3 files changed, 4 insertions(+), 25 deletions(-) diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c index 5d523d5dae88..705c012fcf9e 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -37,6 +37,7 @@ #include #include #include +#include #include #include #include @@ -1574,7 +1575,6 @@ int radeon_suspend_kms(struct drm_device *dev, bool suspend, list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) { struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc); struct drm_framebuffer *fb = crtc->primary->fb; - struct radeon_bo *robj; if (radeon_crtc->cursor_bo) { struct radeon_bo *robj = gem_to_radeon_bo(radeon_crtc->cursor_bo); @@ -1588,9 +1588,10 @@ int radeon_suspend_kms(struct drm_device *dev, bool suspend, if (fb == NULL || fb->obj[0] == NULL) { continue; } - robj = gem_to_radeon_bo(fb->obj[0]); /* don't unpin kernel fb objects */ - if (!radeon_fbdev_robj_is_fb(rdev, robj)) { + if (!drm_fb_helper_gem_is_fb(dev->fb_helper, fb->obj[0])) { + struct radeon_bo *robj = gem_to_radeon_bo(fb->obj[0]); + r = radeon_bo_reserve(robj, false); if (r == 0) { radeon_bo_unpin(robj); diff --git a/drivers/gpu/drm/radeon/radeon_fbdev.c b/drivers/gpu/drm/radeon/radeon_fbdev.c index 18d61f3f7344..3e243f5e2f44 100644 --- a/drivers/gpu/drm/radeon/radeon_fbdev.c +++ b/drivers/gpu/drm/radeon/radeon_fbdev.c @@ -274,20 +274,3 @@ err_radeon_fbdev_destroy_pinned_object: radeon_fbdev_destroy_pinned_object(gobj); return ret; } - -bool radeon_fbdev_robj_is_fb(struct radeon_device *rdev, struct radeon_bo *robj) -{ - struct drm_fb_helper *fb_helper = rdev_to_drm(rdev)->fb_helper; - struct drm_gem_object *gobj; - - if (!fb_helper) - return false; - - gobj = drm_gem_fb_get_obj(fb_helper->fb, 0); - if (!gobj) - return false; - if (gobj != &robj->tbo.base) - return false; - - return true; -} diff --git a/drivers/gpu/drm/radeon/radeon_mode.h b/drivers/gpu/drm/radeon/radeon_mode.h index 088af85902f7..ae1ecdc2e189 100644 --- a/drivers/gpu/drm/radeon/radeon_mode.h +++ b/drivers/gpu/drm/radeon/radeon_mode.h @@ -936,14 +936,9 @@ int radeon_fbdev_driver_fbdev_probe(struct drm_fb_helper *fb_helper, struct drm_fb_helper_surface_size *sizes); #define RADEON_FBDEV_DRIVER_OPS \ .fbdev_probe = radeon_fbdev_driver_fbdev_probe -bool radeon_fbdev_robj_is_fb(struct radeon_device *rdev, struct radeon_bo *robj); #else #define RADEON_FBDEV_DRIVER_OPS \ .fbdev_probe = NULL -static inline bool radeon_fbdev_robj_is_fb(struct radeon_device *rdev, struct radeon_bo *robj) -{ - return false; -} #endif void radeon_crtc_handle_vblank(struct radeon_device *rdev, int crtc_id); -- cgit v1.2.3 From 0b78bb6e6534d3b3bdd849e899fdd17b7f609d04 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Wed, 4 Mar 2026 10:06:04 -0500 Subject: drm/amdgpu: gfx 12.1 cleanups Remove some remnants from when the code was forked from gfx 12.0. Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v12_1.c | 19 ++++--------------- 1 file changed, 4 insertions(+), 15 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v12_1.c b/drivers/gpu/drm/amd/amdgpu/gfx_v12_1.c index 557d15b90ad2..3fd489ea0b79 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v12_1.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v12_1.c @@ -3630,12 +3630,6 @@ static int gfx_v12_1_eop_irq(struct amdgpu_device *adev, return -EINVAL; switch (me_id) { - case 0: - if (pipe_id == 0) - amdgpu_fence_process(&adev->gfx.gfx_ring[0]); - else - amdgpu_fence_process(&adev->gfx.gfx_ring[1]); - break; case 1: case 2: for (i = 0; i < adev->gfx.num_compute_rings; i++) { @@ -3652,6 +3646,9 @@ static int gfx_v12_1_eop_irq(struct amdgpu_device *adev, amdgpu_fence_process(ring); } break; + default: + dev_dbg(adev->dev, "Unexpected me %d in eop_irq\n", me_id); + break; } } @@ -3720,14 +3717,6 @@ static void gfx_v12_1_handle_priv_fault(struct amdgpu_device *adev, return; switch (me_id) { - case 0: - for (i = 0; i < adev->gfx.num_gfx_rings; i++) { - ring = &adev->gfx.gfx_ring[i]; - /* we only enabled 1 gfx queue per pipe for now */ - if (ring->me == me_id && ring->pipe == pipe_id) - drm_sched_fault(&ring->sched); - } - break; case 1: case 2: for (i = 0; i < adev->gfx.num_compute_rings; i++) { @@ -3740,7 +3729,7 @@ static void gfx_v12_1_handle_priv_fault(struct amdgpu_device *adev, } break; default: - BUG(); + dev_dbg(adev->dev, "Unexpected me %d in priv_fault\n", me_id); break; } } -- cgit v1.2.3 From f5a05f8414fc10f307eb965f303580c7778f8dd2 Mon Sep 17 00:00:00 2001 From: Mario Limonciello Date: Thu, 5 Mar 2026 09:06:11 -0600 Subject: drm/amd: Fix a few more NULL pointer dereference in device cleanup I found a few more paths that cleanup fails due to a NULL version pointer on unsupported hardware. Add NULL checks as applicable. Fixes: 39fc2bc4da00 ("drm/amdgpu: Protect GPU register accesses in powergated state in some paths") Reviewed-by: Alex Deucher Signed-off-by: Mario Limonciello Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index f7309705788d..99e5d1e0bf3c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2800,6 +2800,8 @@ static int amdgpu_device_ip_fini_early(struct amdgpu_device *adev) int i, r; for (i = 0; i < adev->num_ip_blocks; i++) { + if (!adev->ip_blocks[i].version) + continue; if (!adev->ip_blocks[i].version->funcs->early_fini) continue; @@ -2877,6 +2879,8 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev) if (!adev->ip_blocks[i].status.sw) continue; + if (!adev->ip_blocks[i].version) + continue; if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) { amdgpu_ucode_free_bo(adev); amdgpu_free_static_csa(&adev->virt.csa_obj); @@ -2903,6 +2907,8 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev) for (i = adev->num_ip_blocks - 1; i >= 0; i--) { if (!adev->ip_blocks[i].status.late_initialized) continue; + if (!adev->ip_blocks[i].version) + continue; if (adev->ip_blocks[i].version->funcs->late_fini) adev->ip_blocks[i].version->funcs->late_fini(&adev->ip_blocks[i]); adev->ip_blocks[i].status.late_initialized = false; -- cgit v1.2.3 From 829f21e90d1982c1b5674828f4e0b9424f345921 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Wed, 28 Jan 2026 14:50:31 -0500 Subject: drm/amdgpu/gfx12.1: add support for disable_kq Plumb in support for disabling kernel queues and make it the default. For testing, kernel queues can be re-enabled by setting amdgpu.user_queue=0 v2: integrate feedback from Lijo Acked-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v12_1.c | 90 +++++++++++++++++++++++++++------- 1 file changed, 71 insertions(+), 19 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v12_1.c b/drivers/gpu/drm/amd/amdgpu/gfx_v12_1.c index 3fd489ea0b79..4e02b62cdbb3 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v12_1.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v12_1.c @@ -1155,11 +1155,13 @@ static int gfx_v12_1_sw_init(struct amdgpu_ip_block *ip_block) break; } - /* recalculate compute rings to use based on hardware configuration */ - num_compute_rings = (adev->gfx.mec.num_pipe_per_mec * - adev->gfx.mec.num_queue_per_pipe) / 2; - adev->gfx.num_compute_rings = min(adev->gfx.num_compute_rings, - num_compute_rings); + if (adev->gfx.num_compute_rings) { + /* recalculate compute rings to use based on hardware configuration */ + num_compute_rings = (adev->gfx.mec.num_pipe_per_mec * + adev->gfx.mec.num_queue_per_pipe) / 2; + adev->gfx.num_compute_rings = min(adev->gfx.num_compute_rings, + num_compute_rings); + } num_xcc = NUM_XCC(adev->gfx.xcc_mask); @@ -2794,6 +2796,33 @@ static void gfx_v12_1_xcc_fini(struct amdgpu_device *adev, gfx_v12_1_xcc_enable_gui_idle_interrupt(adev, false, xcc_id); } +static int gfx_v12_1_set_userq_eop_interrupts(struct amdgpu_device *adev, + bool enable) +{ + unsigned int irq_type; + int m, p, r; + + if (adev->gfx.disable_kq) { + for (m = 0; m < adev->gfx.mec.num_mec; ++m) { + for (p = 0; p < adev->gfx.mec.num_pipe_per_mec; p++) { + irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP + + (m * adev->gfx.mec.num_pipe_per_mec) + + p; + if (enable) + r = amdgpu_irq_get(adev, &adev->gfx.eop_irq, + irq_type); + else + r = amdgpu_irq_put(adev, &adev->gfx.eop_irq, + irq_type); + if (r) + return r; + } + } + } + + return 0; +} + static int gfx_v12_1_hw_fini(struct amdgpu_ip_block *ip_block) { struct amdgpu_device *adev = ip_block->adev; @@ -2801,6 +2830,7 @@ static int gfx_v12_1_hw_fini(struct amdgpu_ip_block *ip_block) amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0); + gfx_v12_1_set_userq_eop_interrupts(adev, false); num_xcc = NUM_XCC(adev->gfx.xcc_mask); for (i = 0; i < num_xcc; i++) { @@ -2868,10 +2898,26 @@ static int gfx_v12_1_early_init(struct amdgpu_ip_block *ip_block) { struct amdgpu_device *adev = ip_block->adev; + + switch (amdgpu_user_queue) { + case -1: + default: + adev->gfx.disable_kq = true; + adev->gfx.disable_uq = true; + break; + case 0: + adev->gfx.disable_kq = false; + adev->gfx.disable_uq = true; + break; + } + adev->gfx.funcs = &gfx_v12_1_gfx_funcs; - adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev), - AMDGPU_MAX_COMPUTE_RINGS); + if (adev->gfx.disable_kq) + adev->gfx.num_compute_rings = 0; + else + adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev), + AMDGPU_MAX_COMPUTE_RINGS); gfx_v12_1_set_kiq_pm4_funcs(adev); gfx_v12_1_set_ring_funcs(adev); @@ -2898,6 +2944,10 @@ static int gfx_v12_1_late_init(struct amdgpu_ip_block *ip_block) if (r) return r; + r = gfx_v12_1_set_userq_eop_interrupts(adev, true); + if (r) + return r; + return 0; } @@ -3716,21 +3766,23 @@ static void gfx_v12_1_handle_priv_fault(struct amdgpu_device *adev, if (xcc_id == -EINVAL) return; - switch (me_id) { - case 1: - case 2: - for (i = 0; i < adev->gfx.num_compute_rings; i++) { - ring = &adev->gfx.compute_ring + if (!adev->gfx.disable_kq) { + switch (me_id) { + case 1: + case 2: + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + ring = &adev->gfx.compute_ring [i + xcc_id * adev->gfx.num_compute_rings]; - if (ring->me == me_id && ring->pipe == pipe_id && - ring->queue == queue_id) - drm_sched_fault(&ring->sched); + if (ring->me == me_id && ring->pipe == pipe_id && + ring->queue == queue_id) + drm_sched_fault(&ring->sched); + } + break; + default: + dev_dbg(adev->dev, "Unexpected me %d in priv_fault\n", me_id); + break; } - break; - default: - dev_dbg(adev->dev, "Unexpected me %d in priv_fault\n", me_id); - break; } } -- cgit v1.2.3 From 9f0e7b5177a452be7a3740128eaa9c23e6660ab9 Mon Sep 17 00:00:00 2001 From: Joshua Peisach Date: Tue, 3 Mar 2026 16:18:22 -0500 Subject: drm/amdgpu/amdgpu_connectors: use struct drm_edid instead of struct edid Some amdgpu code is still using deprecated edid functions. Switch to the newer functions and update the amdgpu_connector struct's edid type to the drm_edid type. At the same time, use the raw EDID when we need to for speaker allocations and for determining if the input is digital. Signed-off-by: Joshua Peisach Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c | 32 +++++++++++++------------- drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h | 2 +- drivers/gpu/drm/amd/amdgpu/dce_v10_0.c | 4 ++-- drivers/gpu/drm/amd/amdgpu/dce_v6_0.c | 4 ++-- drivers/gpu/drm/amd/amdgpu/dce_v8_0.c | 4 ++-- 5 files changed, 23 insertions(+), 23 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c index d1bf2e150c1a..6336cadad164 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c @@ -246,10 +246,10 @@ amdgpu_connector_find_encoder(struct drm_connector *connector, return NULL; } -static struct edid * +static const struct drm_edid * amdgpu_connector_get_hardcoded_edid(struct amdgpu_device *adev) { - return drm_edid_duplicate(drm_edid_raw(adev->mode_info.bios_hardcoded_edid)); + return drm_edid_dup(adev->mode_info.bios_hardcoded_edid); } static void amdgpu_connector_get_edid(struct drm_connector *connector) @@ -268,8 +268,8 @@ static void amdgpu_connector_get_edid(struct drm_connector *connector) if ((amdgpu_connector_encoder_get_dp_bridge_encoder_id(connector) != ENCODER_OBJECT_ID_NONE) && amdgpu_connector->ddc_bus->has_aux) { - amdgpu_connector->edid = drm_get_edid(connector, - &amdgpu_connector->ddc_bus->aux.ddc); + amdgpu_connector->edid = drm_edid_read_ddc(connector, + &amdgpu_connector->ddc_bus->aux.ddc); } else if ((connector->connector_type == DRM_MODE_CONNECTOR_DisplayPort) || (connector->connector_type == DRM_MODE_CONNECTOR_eDP)) { struct amdgpu_connector_atom_dig *dig = amdgpu_connector->con_priv; @@ -277,14 +277,14 @@ static void amdgpu_connector_get_edid(struct drm_connector *connector) if ((dig->dp_sink_type == CONNECTOR_OBJECT_ID_DISPLAYPORT || dig->dp_sink_type == CONNECTOR_OBJECT_ID_eDP) && amdgpu_connector->ddc_bus->has_aux) - amdgpu_connector->edid = drm_get_edid(connector, - &amdgpu_connector->ddc_bus->aux.ddc); + amdgpu_connector->edid = drm_edid_read_ddc(connector, + &amdgpu_connector->ddc_bus->aux.ddc); else if (amdgpu_connector->ddc_bus) - amdgpu_connector->edid = drm_get_edid(connector, - &amdgpu_connector->ddc_bus->adapter); + amdgpu_connector->edid = drm_edid_read_ddc(connector, + &amdgpu_connector->ddc_bus->adapter); } else if (amdgpu_connector->ddc_bus) { - amdgpu_connector->edid = drm_get_edid(connector, - &amdgpu_connector->ddc_bus->adapter); + amdgpu_connector->edid = drm_edid_read_ddc(connector, + &amdgpu_connector->ddc_bus->adapter); } if (!amdgpu_connector->edid) { @@ -292,7 +292,7 @@ static void amdgpu_connector_get_edid(struct drm_connector *connector) if (((connector->connector_type == DRM_MODE_CONNECTOR_LVDS) || (connector->connector_type == DRM_MODE_CONNECTOR_eDP))) { amdgpu_connector->edid = amdgpu_connector_get_hardcoded_edid(adev); - drm_connector_update_edid_property(connector, amdgpu_connector->edid); + drm_edid_connector_update(connector, amdgpu_connector->edid); } } } @@ -311,11 +311,11 @@ static int amdgpu_connector_ddc_get_modes(struct drm_connector *connector) int ret; if (amdgpu_connector->edid) { - drm_connector_update_edid_property(connector, amdgpu_connector->edid); - ret = drm_add_edid_modes(connector, amdgpu_connector->edid); + drm_edid_connector_update(connector, amdgpu_connector->edid); + ret = drm_edid_connector_add_modes(connector); return ret; } - drm_connector_update_edid_property(connector, NULL); + drm_edid_connector_update(connector, NULL); return 0; } @@ -883,7 +883,7 @@ amdgpu_connector_vga_detect(struct drm_connector *connector, bool force) ret = connector_status_connected; } else { amdgpu_connector->use_digital = - !!(amdgpu_connector->edid->input & DRM_EDID_INPUT_DIGITAL); + drm_edid_is_digital(amdgpu_connector->edid); /* some oems have boards with separate digital and analog connectors * with a shared ddc line (often vga + hdmi) @@ -1063,7 +1063,7 @@ amdgpu_connector_dvi_detect(struct drm_connector *connector, bool force) broken_edid = true; /* defer use_digital to later */ } else { amdgpu_connector->use_digital = - !!(amdgpu_connector->edid->input & DRM_EDID_INPUT_DIGITAL); + drm_edid_is_digital(amdgpu_connector->edid); /* some oems have boards with separate digital and analog connectors * with a shared ddc line (often vga + hdmi) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h index dc8d2f52c7d6..c4e0255814e0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h @@ -624,7 +624,7 @@ struct amdgpu_connector { bool use_digital; /* we need to mind the EDID between detect and get modes due to analog/digital/tvencoder */ - struct edid *edid; + const struct drm_edid *edid; void *con_priv; bool dac_load_detect; bool detected_by_load; /* if the connection status was determined by load */ diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c index f1052acea5ec..c8f465158e71 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c @@ -1298,7 +1298,7 @@ static void dce_v10_0_audio_write_speaker_allocation(struct drm_encoder *encoder return; } - sad_count = drm_edid_to_speaker_allocation(amdgpu_connector->edid, &sadb); + sad_count = drm_edid_to_speaker_allocation(drm_edid_raw(amdgpu_connector->edid), &sadb); if (sad_count < 0) { DRM_ERROR("Couldn't read Speaker Allocation Data Block: %d\n", sad_count); sad_count = 0; @@ -1368,7 +1368,7 @@ static void dce_v10_0_audio_write_sad_regs(struct drm_encoder *encoder) return; } - sad_count = drm_edid_to_sad(amdgpu_connector->edid, &sads); + sad_count = drm_edid_to_sad(drm_edid_raw(amdgpu_connector->edid), &sads); if (sad_count < 0) DRM_ERROR("Couldn't read SADs: %d\n", sad_count); if (sad_count <= 0) diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c index c153a6e1e22a..58d0da5c2a74 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c @@ -1265,7 +1265,7 @@ static void dce_v6_0_audio_write_speaker_allocation(struct drm_encoder *encoder) return; } - sad_count = drm_edid_to_speaker_allocation(amdgpu_connector->edid, &sadb); + sad_count = drm_edid_to_speaker_allocation(drm_edid_raw(amdgpu_connector->edid), &sadb); if (sad_count < 0) { DRM_ERROR("Couldn't read Speaker Allocation Data Block: %d\n", sad_count); sad_count = 0; @@ -1346,7 +1346,7 @@ static void dce_v6_0_audio_write_sad_regs(struct drm_encoder *encoder) return; } - sad_count = drm_edid_to_sad(amdgpu_connector->edid, &sads); + sad_count = drm_edid_to_sad(drm_edid_raw(amdgpu_connector->edid), &sads); if (sad_count < 0) DRM_ERROR("Couldn't read SADs: %d\n", sad_count); if (sad_count <= 0) diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c index a85a9e32fde4..6d19f6d94d25 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c @@ -1271,7 +1271,7 @@ static void dce_v8_0_audio_write_speaker_allocation(struct drm_encoder *encoder) return; } - sad_count = drm_edid_to_speaker_allocation(amdgpu_connector->edid, &sadb); + sad_count = drm_edid_to_speaker_allocation(drm_edid_raw(amdgpu_connector->edid), &sadb); if (sad_count < 0) { DRM_ERROR("Couldn't read Speaker Allocation Data Block: %d\n", sad_count); sad_count = 0; @@ -1339,7 +1339,7 @@ static void dce_v8_0_audio_write_sad_regs(struct drm_encoder *encoder) return; } - sad_count = drm_edid_to_sad(amdgpu_connector->edid, &sads); + sad_count = drm_edid_to_sad(drm_edid_raw(amdgpu_connector->edid), &sads); if (sad_count < 0) DRM_ERROR("Couldn't read SADs: %d\n", sad_count); if (sad_count <= 0) -- cgit v1.2.3 From 71036457ad85a5b08e5101738c4803a45b4e1555 Mon Sep 17 00:00:00 2001 From: Joshua Peisach Date: Tue, 3 Mar 2026 16:18:23 -0500 Subject: drm/amdgpu/amdgpu_connectors: remove amdgpu_connector_free_edid Now that we are using struct drm_edid, we can just call drm_edid_free directly. Remove the function and update calls to drm_edid_free. Signed-off-by: Joshua Peisach Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c | 22 +++++++--------------- 1 file changed, 7 insertions(+), 15 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c index 6336cadad164..aabe9d58c3dc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c @@ -297,14 +297,6 @@ static void amdgpu_connector_get_edid(struct drm_connector *connector) } } -static void amdgpu_connector_free_edid(struct drm_connector *connector) -{ - struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector); - - kfree(amdgpu_connector->edid); - amdgpu_connector->edid = NULL; -} - static int amdgpu_connector_ddc_get_modes(struct drm_connector *connector) { struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector); @@ -754,7 +746,7 @@ static void amdgpu_connector_destroy(struct drm_connector *connector) { struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector); - amdgpu_connector_free_edid(connector); + drm_edid_free(amdgpu_connector->edid); kfree(amdgpu_connector->con_priv); drm_connector_unregister(connector); drm_connector_cleanup(connector); @@ -873,7 +865,7 @@ amdgpu_connector_vga_detect(struct drm_connector *connector, bool force) dret = amdgpu_display_ddc_probe(amdgpu_connector, false); if (dret) { amdgpu_connector->detected_by_load = false; - amdgpu_connector_free_edid(connector); + drm_edid_free(amdgpu_connector->edid); amdgpu_connector_get_edid(connector); if (!amdgpu_connector->edid) { @@ -889,7 +881,7 @@ amdgpu_connector_vga_detect(struct drm_connector *connector, bool force) * with a shared ddc line (often vga + hdmi) */ if (amdgpu_connector->use_digital && amdgpu_connector->shared_ddc) { - amdgpu_connector_free_edid(connector); + drm_edid_free(amdgpu_connector->edid); ret = connector_status_disconnected; } else { ret = connector_status_connected; @@ -984,7 +976,7 @@ static void amdgpu_connector_shared_ddc(enum drm_connector_status *status, /* hpd is our only option in this case */ if (!amdgpu_display_hpd_sense(adev, amdgpu_connector->hpd.hpd)) { - amdgpu_connector_free_edid(connector); + drm_edid_free(amdgpu_connector->edid); *status = connector_status_disconnected; } } @@ -1053,7 +1045,7 @@ amdgpu_connector_dvi_detect(struct drm_connector *connector, bool force) } if (dret) { amdgpu_connector->detected_by_load = false; - amdgpu_connector_free_edid(connector); + drm_edid_free(amdgpu_connector->edid); amdgpu_connector_get_edid(connector); if (!amdgpu_connector->edid) { @@ -1069,7 +1061,7 @@ amdgpu_connector_dvi_detect(struct drm_connector *connector, bool force) * with a shared ddc line (often vga + hdmi) */ if ((!amdgpu_connector->use_digital) && amdgpu_connector->shared_ddc) { - amdgpu_connector_free_edid(connector); + drm_edid_free(amdgpu_connector->edid); ret = connector_status_disconnected; } else { ret = connector_status_connected; @@ -1417,7 +1409,7 @@ amdgpu_connector_dp_detect(struct drm_connector *connector, bool force) goto out; } - amdgpu_connector_free_edid(connector); + drm_edid_free(amdgpu_connector->edid); if ((connector->connector_type == DRM_MODE_CONNECTOR_eDP) || (connector->connector_type == DRM_MODE_CONNECTOR_LVDS)) { -- cgit v1.2.3 From ec5708d6e547f7efe2f009073bfa98dbc4c5c2ac Mon Sep 17 00:00:00 2001 From: Cristian Ciocaltea Date: Thu, 5 Mar 2026 13:16:36 +0200 Subject: drm/amdgpu: Fix kernel-doc comments for some LUT properties The following members of struct amdgpu_mode_info do not have valid references in the related kernel-doc sections: - plane_shaper_lut_property - plane_shaper_lut_size_property, - plane_lut3d_size_property Correct all affected comment blocks. Fixes: f545d82479b4 ("drm/amd/display: add plane shaper LUT and TF driver-specific properties") Fixes: 671994e3bf33 ("drm/amd/display: add plane 3D LUT driver-specific properties") Reviewed-by: Melissa Wen Signed-off-by: Cristian Ciocaltea Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h index c4e0255814e0..90352284c5ee 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h @@ -368,15 +368,15 @@ struct amdgpu_mode_info { struct drm_property *plane_ctm_property; /** - * @shaper_lut_property: Plane property to set pre-blending shaper LUT - * that converts color content before 3D LUT. If - * plane_shaper_tf_property != Identity TF, AMD color module will + * @plane_shaper_lut_property: Plane property to set pre-blending + * shaper LUT that converts color content before 3D LUT. + * If plane_shaper_tf_property != Identity TF, AMD color module will * combine the user LUT values with pre-defined TF into the LUT * parameters to be programmed. */ struct drm_property *plane_shaper_lut_property; /** - * @shaper_lut_size_property: Plane property for the size of + * @plane_shaper_lut_size_property: Plane property for the size of * pre-blending shaper LUT as supported by the driver (read-only). */ struct drm_property *plane_shaper_lut_size_property; @@ -400,10 +400,10 @@ struct amdgpu_mode_info { */ struct drm_property *plane_lut3d_property; /** - * @plane_degamma_lut_size_property: Plane property to define the max - * size of 3D LUT as supported by the driver (read-only). The max size - * is the max size of one dimension and, therefore, the max number of - * entries for 3D LUT array is the 3D LUT size cubed; + * @plane_lut3d_size_property: Plane property to define the max size + * of 3D LUT as supported by the driver (read-only). The max size is + * the max size of one dimension and, therefore, the max number of + * entries for 3D LUT array is the 3D LUT size cubed. */ struct drm_property *plane_lut3d_size_property; /** -- cgit v1.2.3 From 4e22a5fe6ea6e0b057e7f246df4ac3ff8bfbc46a Mon Sep 17 00:00:00 2001 From: Yang Wang Date: Wed, 4 Mar 2026 18:45:45 -0500 Subject: drm/amdgpu: fix gpu idle power consumption issue for gfx v12 Older versions of the MES firmware may cause abnormal GPU power consumption. When performing inference tasks on the GPU (e.g., with Ollama using ROCm), the GPU may show abnormal power consumption in idle state and incorrect GPU load information. This issue has been fixed in firmware version 0x8b and newer. Closes: https://github.com/ROCm/ROCm/issues/5706 Signed-off-by: Yang Wang Acked-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/mes_v12_0.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c b/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c index 5bfa5d1d0b36..023c7345ea54 100644 --- a/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c @@ -731,6 +731,9 @@ static int mes_v12_0_set_hw_resources(struct amdgpu_mes *mes, int pipe) int i; struct amdgpu_device *adev = mes->adev; union MESAPI_SET_HW_RESOURCES mes_set_hw_res_pkt; + uint32_t mes_rev = (pipe == AMDGPU_MES_SCHED_PIPE) ? + (mes->sched_version & AMDGPU_MES_VERSION_MASK) : + (mes->kiq_version & AMDGPU_MES_VERSION_MASK); memset(&mes_set_hw_res_pkt, 0, sizeof(mes_set_hw_res_pkt)); @@ -785,7 +788,7 @@ static int mes_v12_0_set_hw_resources(struct amdgpu_mes *mes, int pipe) * handling support, other queue will not use the oversubscribe timer. * handling mode - 0: disabled; 1: basic version; 2: basic+ version */ - mes_set_hw_res_pkt.oversubscription_timer = 50; + mes_set_hw_res_pkt.oversubscription_timer = mes_rev < 0x8b ? 0 : 50; mes_set_hw_res_pkt.unmapped_doorbell_handling = 1; if (amdgpu_mes_log_enable) { -- cgit v1.2.3 From e81eff80aad6c04ff1d7c1bf6412de17e717bd76 Mon Sep 17 00:00:00 2001 From: Pierre-Eric Pelloux-Prayer Date: Tue, 2 Dec 2025 14:27:24 +0100 Subject: drm/amdgpu: include ip discovery data in devcoredump This is the best way to describe the GPU to a tool loading the devcoredump. Signed-off-by: Pierre-Eric Pelloux-Prayer Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c | 2 ++ drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 42 ++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h | 3 ++ 3 files changed, 47 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c index c38e7371bafc..160f0704d1d3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c @@ -261,6 +261,8 @@ amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count, } } + amdgpu_discovery_dump(coredump->adev, &p); + /* IP firmware information */ drm_printf(&p, "\nIP Firmwares\n"); amdgpu_devcoredump_fw_info(coredump->adev, &p); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index 6c8b3c2687dc..c4ae1ce52a0f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -1379,6 +1379,48 @@ static void amdgpu_discovery_sysfs_fini(struct amdgpu_device *adev) kobject_put(&ip_top->kobj); } +/* devcoredump support */ +void amdgpu_discovery_dump(struct amdgpu_device *adev, struct drm_printer *p) +{ + struct ip_discovery_top *ip_top = adev->discovery.ip_top; + struct ip_die_entry *ip_die_entry; + struct list_head *el_die, *el_hw_id, *el_hw_inst; + struct ip_hw_id *hw_id; + struct kset *die_kset; + struct ip_hw_instance *ip_inst; + int i = 0, j; + + die_kset = &ip_top->die_kset; + + drm_printf(p, "\nHW IP Discovery\n"); + spin_lock(&die_kset->list_lock); + list_for_each(el_die, &die_kset->list) { + drm_printf(p, "die %d\n", i++); + ip_die_entry = to_ip_die_entry(list_to_kobj(el_die)); + + list_for_each(el_hw_id, &ip_die_entry->ip_kset.list) { + hw_id = to_ip_hw_id(list_to_kobj(el_hw_id)); + drm_printf(p, "hw_id %d %s\n", hw_id->hw_id, hw_id_names[hw_id->hw_id]); + + list_for_each(el_hw_inst, &hw_id->hw_id_kset.list) { + ip_inst = to_ip_hw_instance(list_to_kobj(el_hw_inst)); + drm_printf(p, "\tinstance %d\n", ip_inst->num_instance); + drm_printf(p, "\tmajor %d\n", ip_inst->major); + drm_printf(p, "\tminor %d\n", ip_inst->minor); + drm_printf(p, "\trevision %d\n", ip_inst->revision); + drm_printf(p, "\tharvest 0x%01X\n", ip_inst->harvest); + drm_printf(p, "\tnum_base_addresses %d\n", + ip_inst->num_base_addresses); + for (j = 0; j < ip_inst->num_base_addresses; j++) + drm_printf(p, "\tbase_addr[%d] 0x%08X\n", + j, ip_inst->base_addr[j]); + } + } + } + spin_unlock(&die_kset->list_lock); +} + + /* ================================================== */ static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h index 4ce04486cc31..c8242992c912 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h @@ -30,6 +30,7 @@ #define DISCOVERY_TMR_OFFSET (64 << 10) struct ip_discovery_top; +struct drm_printer; struct amdgpu_discovery_info { struct debugfs_blob_wrapper debugfs_blob; @@ -47,4 +48,6 @@ int amdgpu_discovery_get_nps_info(struct amdgpu_device *adev, struct amdgpu_gmc_memrange **ranges, int *range_cnt, bool refresh); +void amdgpu_discovery_dump(struct amdgpu_device *adev, struct drm_printer *p); + #endif /* __AMDGPU_DISCOVERY__ */ -- cgit v1.2.3 From 281108ad1e2d75216e29d8c9877a7bb32839659d Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Mon, 2 Mar 2026 18:26:55 +0800 Subject: drm/amd/pm: Add aid/xcd/hbm temperatures Add aid, xcd & hbm temperatures to gpu metrics for smu_v13_0_12 v2: Use correct umc control per stack (Lijo) Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/include/kgd_pp_interface.h | 3 ++ .../gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c | 37 +++++++++++++++++++++- .../gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h | 13 +++++++- 3 files changed, 51 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/include/kgd_pp_interface.h b/drivers/gpu/drm/amd/include/kgd_pp_interface.h index bdf8e6ff556c..a9b73f4fd466 100644 --- a/drivers/gpu/drm/amd/include/kgd_pp_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_pp_interface.h @@ -584,6 +584,9 @@ enum amdgpu_metrics_attr_id { AMDGPU_METRICS_ATTR_ID_GFX_BELOW_HOST_LIMIT_THM_ACC, AMDGPU_METRICS_ATTR_ID_GFX_LOW_UTILIZATION_ACC, AMDGPU_METRICS_ATTR_ID_GFX_BELOW_HOST_LIMIT_TOTAL_ACC, + AMDGPU_METRICS_ATTR_ID_TEMPERATURE_HBM, + AMDGPU_METRICS_ATTR_ID_TEMPERATURE_AID, + AMDGPU_METRICS_ATTR_ID_TEMPERATURE_XCD, AMDGPU_METRICS_ATTR_ID_MAX, }; diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c index 32d5e2170d80..8610f3da514f 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c @@ -49,6 +49,13 @@ #undef pr_info #undef pr_debug +#define hbm_stack_mask_valid(umc_mask) \ + (((umc_mask) & 0x3) == 0x3) + +#define for_each_hbm_stack(stack_idx, umc_mask) \ + for ((stack_idx) = 0; (umc_mask); \ + (umc_mask) >>= 2, (stack_idx)++) \ + #define SMU_13_0_12_FEA_MAP(smu_feature, smu_13_0_12_feature) \ [smu_feature] = { 1, (smu_13_0_12_feature) } @@ -834,7 +841,7 @@ void smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table, struct smu_v13_0_6_gpu_metrics *gpu_metrics) { struct amdgpu_device *adev = smu->adev; - int ret = 0, xcc_id, inst, i, j; + int ret = 0, xcc_id, inst, i, j, idx; u8 num_jpeg_rings_gpu_metrics; MetricsTable_t *metrics; @@ -849,6 +856,31 @@ void smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table, gpu_metrics->temperature_vrsoc = SMUQ10_ROUND(metrics->MaxVrTemperature); + if (smu_v13_0_6_cap_supported(smu, + SMU_CAP(TEMP_AID_XCD_HBM))) { + if (adev->umc.active_mask) { + u64 mask = adev->umc.active_mask; + int out_idx = 0; + int stack_idx; + + if (unlikely(hweight64(mask) / 2 > SMU_13_0_6_MAX_HBM_STACKS)) { + dev_warn(adev->dev, "Invalid umc mask %lld\n", mask); + } else { + for_each_hbm_stack(stack_idx, mask) { + if (!hbm_stack_mask_valid(mask)) + continue; + gpu_metrics->temperature_hbm[out_idx++] = + metrics->HbmTemperature[stack_idx]; + } + } + } + idx = 0; + for_each_inst(i, adev->aid_mask) { + gpu_metrics->temperature_aid[idx] = metrics->AidTemperature[i]; + idx++; + } + } + gpu_metrics->average_gfx_activity = SMUQ10_ROUND(metrics->SocketGfxBusy); gpu_metrics->average_umc_activity = @@ -964,6 +996,9 @@ void smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table, [i] = SMUQ10_ROUND( metrics->GfxclkBelowHostLimitTotalAcc[inst]); } + if (smu_v13_0_6_cap_supported(smu, + SMU_CAP(TEMP_AID_XCD_HBM))) + gpu_metrics->temperature_xcd[i] = metrics->XcdTemperature[inst]; } gpu_metrics->xgmi_link_width = metrics->XgmiWidth; diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h index ffb06564f830..a150fc88902c 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h @@ -78,6 +78,7 @@ enum smu_v13_0_6_caps { SMU_CAP(RAS_EEPROM), SMU_CAP(FAST_PPT), SMU_CAP(SYSTEM_POWER_METRICS), + SMU_CAP(TEMP_AID_XCD_HBM), SMU_CAP(ALL), }; @@ -87,6 +88,8 @@ enum smu_v13_0_6_caps { #define SMU_13_0_6_MAX_XCC 8 #define SMU_13_0_6_MAX_VCN 4 #define SMU_13_0_6_MAX_JPEG 40 +#define SMU_13_0_6_MAX_AID 4 +#define SMU_13_0_6_MAX_HBM_STACKS 8 extern void smu_v13_0_6_set_ppt_funcs(struct smu_context *smu); bool smu_v13_0_6_cap_supported(struct smu_context *smu, enum smu_v13_0_6_caps cap); @@ -222,7 +225,15 @@ extern const struct ras_smu_drv smu_v13_0_12_ras_smu_drv; SMU_13_0_6_MAX_XCC); \ SMU_ARRAY(SMU_MATTR(GFX_BELOW_HOST_LIMIT_TOTAL_ACC), SMU_MUNIT(NONE), \ SMU_MTYPE(U64), gfx_below_host_limit_total_acc, \ - SMU_13_0_6_MAX_XCC); + SMU_13_0_6_MAX_XCC); \ + SMU_ARRAY(SMU_MATTR(TEMPERATURE_HBM), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(U16), temperature_hbm, \ + SMU_13_0_6_MAX_HBM_STACKS); \ + SMU_ARRAY(SMU_MATTR(TEMPERATURE_AID), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(U16), temperature_aid, SMU_13_0_6_MAX_AID); \ + SMU_ARRAY(SMU_MATTR(TEMPERATURE_XCD), SMU_MUNIT(TEMP_1), \ + SMU_MTYPE(U16), temperature_xcd, SMU_13_0_6_MAX_XCC); \ + DECLARE_SMU_METRICS_CLASS(smu_v13_0_6_gpu_metrics, SMU_13_0_6_METRICS_FIELDS); void smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table, -- cgit v1.2.3 From 79b7209c9c61584e8584f8bfe8d259e202fe7107 Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Tue, 3 Mar 2026 15:11:00 +0800 Subject: drm/amd/pm: Enable aid/xcd/hbm temperature reporting Enable aid/xcd/hbm temperature reporting for smu_v13_0_12 via gpu metrics Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c index 2512a8ff6836..8dc8674b7ce1 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c @@ -373,6 +373,9 @@ static void smu_v13_0_12_init_caps(struct smu_context *smu) } else { smu_v13_0_12_tables_fini(smu); } + + if (fw_ver >= 0x04561000) + smu_v13_0_6_cap_set(smu, SMU_CAP(TEMP_AID_XCD_HBM)); } static void smu_v13_0_6_init_caps(struct smu_context *smu) -- cgit v1.2.3 From 210fa907a64496af6341c3455cb267d2b68dfb12 Mon Sep 17 00:00:00 2001 From: Sunil Khatri Date: Fri, 6 Mar 2026 16:48:02 +0530 Subject: drm/amdgpu: fix warning for potentially insecure string MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit linux/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c:2358:24: warning: format string is not a string literal (potentially insecure) [-Wformat-security] sprintf(ring->name, amdgpu_sw_ring_name(i)); ^~~~~~~~~~~~~~~~~~~~~~ linux/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c:2358:24: note: treat the string as an argument to avoid this sprintf(ring->name, amdgpu_sw_ring_name(i)); Signed-off-by: Sunil Khatri Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 7e9d753f4a80..95be105671ec 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -2355,7 +2355,7 @@ static int gfx_v9_0_sw_init(struct amdgpu_ip_block *ip_block) for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++) { ring = &adev->gfx.sw_gfx_ring[i]; ring->ring_obj = NULL; - sprintf(ring->name, amdgpu_sw_ring_name(i)); + strscpy(ring->name, amdgpu_sw_ring_name(i), sizeof(ring->name)); ring->use_doorbell = true; ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1; ring->is_sw_ring = true; -- cgit v1.2.3 From 8c78845bf9b18d336e5b5d00c966176a3c3f9581 Mon Sep 17 00:00:00 2001 From: Sunil Khatri Date: Fri, 6 Mar 2026 12:53:30 +0530 Subject: drm/amdkfd: fix the warning for potential insecure string MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Below is the warning thrown by the clang compiler: linux/drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c:588:9: warning: format string is not a string literal (potentially insecure) [-Wformat-security] stats_dir_filename); ^~~~~~~~~~~~~~~~~~ linux/drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c:588:9: note: treat the string as an argument to avoid this stats_dir_filename); ^ "%s", linux/drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c:635:18: warning: format string is not a string literal (potentially insecure) [-Wformat-security] p->kobj, counters_dir_filename); ^~~~~~~~~~~~~~~~~~~~~ linux/drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_process.c:635:18: note: treat the string as an argument to avoid this p->kobj, counters_dir_filename); ^ "%s", Signed-off-by: Sunil Khatri CC: Philip Yang CC: Felix Kuehling Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdkfd/kfd_process.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 12e24fbf8c46..a031166f270c 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -585,7 +585,7 @@ static void kfd_procfs_add_sysfs_stats(struct kfd_process *p) ret = kobject_init_and_add(pdd->kobj_stats, &procfs_stats_type, p->kobj, - stats_dir_filename); + "%s", stats_dir_filename); if (ret) { pr_warn("Creating KFD proc/stats_%s folder failed", @@ -632,7 +632,7 @@ static void kfd_procfs_add_sysfs_counters(struct kfd_process *p) return; ret = kobject_init_and_add(kobj_counters, &sysfs_counters_type, - p->kobj, counters_dir_filename); + p->kobj, "%s", counters_dir_filename); if (ret) { pr_warn("Creating KFD proc/%s folder failed", counters_dir_filename); -- cgit v1.2.3 From 421c0f19043337a553e802b2dfe4b56d538ef4d6 Mon Sep 17 00:00:00 2001 From: David Francis Date: Fri, 16 Jan 2026 10:21:15 -0500 Subject: drm/amdgpu: Check for multiplication overflow in checkpoint stack size get_checkpoint_info() in kfd_mqd_manager_v9.c finds 32-bit value ctl_stack_size by multiplying two 32-bit values. This can overflow to a lower value, which could result in copying outside the bounds of a buffer in checkpoint_mqd() in the same file. Put in a check for the overflow, and fail with -EINVAL if detected. v2: use check_mul_overflow() Signed-off-by: David Francis Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 7 +++++-- drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h | 2 +- drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h | 3 ++- drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c | 7 +++++-- drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c | 3 ++- drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 8 +++++++- 6 files changed, 22 insertions(+), 8 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index 3ddf06c755b5..ab3b2e7be9bd 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -2720,7 +2720,7 @@ static int get_wave_state(struct device_queue_manager *dqm, ctl_stack, ctl_stack_used_size, save_area_used_size); } -static void get_queue_checkpoint_info(struct device_queue_manager *dqm, +static int get_queue_checkpoint_info(struct device_queue_manager *dqm, const struct queue *q, u32 *mqd_size, u32 *ctl_stack_size) @@ -2728,6 +2728,7 @@ static void get_queue_checkpoint_info(struct device_queue_manager *dqm, struct mqd_manager *mqd_mgr; enum KFD_MQD_TYPE mqd_type = get_mqd_type_from_queue_type(q->properties.type); + int ret = 0; dqm_lock(dqm); mqd_mgr = dqm->mqd_mgrs[mqd_type]; @@ -2735,9 +2736,11 @@ static void get_queue_checkpoint_info(struct device_queue_manager *dqm, *ctl_stack_size = 0; if (q->properties.type == KFD_QUEUE_TYPE_COMPUTE && mqd_mgr->get_checkpoint_info) - mqd_mgr->get_checkpoint_info(mqd_mgr, q->mqd, ctl_stack_size); + ret = mqd_mgr->get_checkpoint_info(mqd_mgr, q->mqd, ctl_stack_size); dqm_unlock(dqm); + + return ret; } static int checkpoint_mqd(struct device_queue_manager *dqm, diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h index ef07e44916f8..3272328da11f 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h @@ -192,7 +192,7 @@ struct device_queue_manager_ops { int (*reset_queues)(struct device_queue_manager *dqm, uint16_t pasid); - void (*get_queue_checkpoint_info)(struct device_queue_manager *dqm, + int (*get_queue_checkpoint_info)(struct device_queue_manager *dqm, const struct queue *q, u32 *mqd_size, u32 *ctl_stack_size); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h index 2429d278ef0e..06ca6235ff1b 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h @@ -102,7 +102,8 @@ struct mqd_manager { u32 *ctl_stack_used_size, u32 *save_area_used_size); - void (*get_checkpoint_info)(struct mqd_manager *mm, void *mqd, uint32_t *ctl_stack_size); + int (*get_checkpoint_info)(struct mqd_manager *mm, void *mqd, + uint32_t *ctl_stack_size); void (*checkpoint_mqd)(struct mqd_manager *mm, void *mqd, diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c index 19f21932a5ce..979ae94ac966 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c @@ -385,11 +385,14 @@ static int get_wave_state(struct mqd_manager *mm, void *mqd, return 0; } -static void get_checkpoint_info(struct mqd_manager *mm, void *mqd, u32 *ctl_stack_size) +static int get_checkpoint_info(struct mqd_manager *mm, void *mqd, u32 *ctl_stack_size) { struct v9_mqd *m = get_mqd(mqd); - *ctl_stack_size = m->cp_hqd_cntl_stack_size * NUM_XCC(mm->dev->xcc_mask); + if (check_mul_overflow(m->cp_hqd_cntl_stack_size, NUM_XCC(mm->dev->xcc_mask), ctl_stack_size)) + return -EINVAL; + + return 0; } static void checkpoint_mqd(struct mqd_manager *mm, void *mqd, void *mqd_dst, void *ctl_stack_dst) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c index f02ef2d44a07..431a20323146 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c @@ -274,10 +274,11 @@ static int get_wave_state(struct mqd_manager *mm, void *mqd, return 0; } -static void get_checkpoint_info(struct mqd_manager *mm, void *mqd, u32 *ctl_stack_size) +static int get_checkpoint_info(struct mqd_manager *mm, void *mqd, u32 *ctl_stack_size) { /* Control stack is stored in user mode */ *ctl_stack_size = 0; + return 0; } static void checkpoint_mqd(struct mqd_manager *mm, void *mqd, void *mqd_dst, void *ctl_stack_dst) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c index 8ea31699d38b..586d409ebe4e 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c @@ -1069,6 +1069,7 @@ int pqm_get_queue_checkpoint_info(struct process_queue_manager *pqm, uint32_t *ctl_stack_size) { struct process_queue_node *pqn; + int ret; pqn = get_queue_by_qid(pqm, qid); if (!pqn) { @@ -1081,9 +1082,14 @@ int pqm_get_queue_checkpoint_info(struct process_queue_manager *pqm, return -EOPNOTSUPP; } - pqn->q->device->dqm->ops.get_queue_checkpoint_info(pqn->q->device->dqm, + ret = pqn->q->device->dqm->ops.get_queue_checkpoint_info(pqn->q->device->dqm, pqn->q, mqd_size, ctl_stack_size); + if (ret) { + pr_debug("amdkfd: Overflow while computing stack size for queue %d\n", qid); + return ret; + } + return 0; } -- cgit v1.2.3 From 28cacaace5cde8318b7da967b3955a73cc6de91a Mon Sep 17 00:00:00 2001 From: Sunil Khatri Date: Wed, 4 Mar 2026 00:38:06 +0530 Subject: drm/amdgpu/userq: defer queue publication until create completes MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The userq create path publishes queues to global xarrays such as userq_doorbell_xa and userq_xa before creation was fully complete. Later on if create queue fails, teardown could free an already visible queue, opening a UAF race with concurrent queue walkers. Also calling amdgpu_userq_put in such cases complicates the cleanup. Solution is to defer queue publication until create succeeds and no partially initialized queue is exposed. Signed-off-by: Sunil Khatri Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 66 +++++++++++++++---------------- 1 file changed, 33 insertions(+), 33 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c index 3ae7d1f5d115..049227fdb73a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c @@ -758,7 +758,6 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args) const struct amdgpu_userq_funcs *uq_funcs; struct amdgpu_usermode_queue *queue; struct amdgpu_db_info db_info; - char *queue_name; bool skip_map_queue; u32 qid; uint64_t index; @@ -848,32 +847,6 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args) goto unlock; } - /* drop this refcount during queue destroy */ - kref_init(&queue->refcount); - - /* Wait for mode-1 reset to complete */ - down_read(&adev->reset_domain->sem); - r = xa_err(xa_store_irq(&adev->userq_doorbell_xa, index, queue, GFP_KERNEL)); - if (r) { - kfree(queue); - up_read(&adev->reset_domain->sem); - goto unlock; - } - - r = xa_alloc(&uq_mgr->userq_xa, &qid, queue, - XA_LIMIT(1, AMDGPU_MAX_USERQ_COUNT), GFP_KERNEL); - if (r) { - drm_file_err(uq_mgr->file, "Failed to allocate a queue id\n"); - amdgpu_userq_fence_driver_free(queue); - xa_erase_irq(&adev->userq_doorbell_xa, index); - uq_funcs->mqd_destroy(queue); - kfree(queue); - r = -ENOMEM; - up_read(&adev->reset_domain->sem); - goto unlock; - } - up_read(&adev->reset_domain->sem); - /* don't map the queue if scheduling is halted */ if (adev->userq_halt_for_enforce_isolation && ((queue->queue_type == AMDGPU_HW_IP_GFX) || @@ -885,28 +858,55 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args) r = amdgpu_userq_map_helper(queue); if (r) { drm_file_err(uq_mgr->file, "Failed to map Queue\n"); - xa_erase_irq(&adev->userq_doorbell_xa, index); - xa_erase(&uq_mgr->userq_xa, qid); - amdgpu_userq_fence_driver_free(queue); uq_funcs->mqd_destroy(queue); + amdgpu_userq_fence_driver_free(queue); kfree(queue); goto unlock; } } - queue_name = kasprintf(GFP_KERNEL, "queue-%d", qid); - if (!queue_name) { + /* drop this refcount during queue destroy */ + kref_init(&queue->refcount); + + /* Wait for mode-1 reset to complete */ + down_read(&adev->reset_domain->sem); + r = xa_alloc(&uq_mgr->userq_xa, &qid, queue, + XA_LIMIT(1, AMDGPU_MAX_USERQ_COUNT), GFP_KERNEL); + if (r) { + if (!skip_map_queue) + amdgpu_userq_unmap_helper(queue); + + uq_funcs->mqd_destroy(queue); + amdgpu_userq_fence_driver_free(queue); + kfree(queue); r = -ENOMEM; + up_read(&adev->reset_domain->sem); goto unlock; } + r = xa_err(xa_store_irq(&adev->userq_doorbell_xa, index, queue, GFP_KERNEL)); + if (r) { + xa_erase(&uq_mgr->userq_xa, qid); + if (!skip_map_queue) + amdgpu_userq_unmap_helper(queue); + + uq_funcs->mqd_destroy(queue); + amdgpu_userq_fence_driver_free(queue); + kfree(queue); + up_read(&adev->reset_domain->sem); + goto unlock; + } + up_read(&adev->reset_domain->sem); + #if defined(CONFIG_DEBUG_FS) + char queue_name[32]; + + scnprintf(queue_name, sizeof(queue_name), "queue_%d", qid); /* Queue dentry per client to hold MQD information */ queue->debugfs_queue = debugfs_create_dir(queue_name, filp->debugfs_client); debugfs_create_file("mqd_info", 0444, queue->debugfs_queue, queue, &amdgpu_mqd_info_fops); #endif amdgpu_userq_init_hang_detect_work(queue); - kfree(queue_name); args->out.queue_id = qid; atomic_inc(&uq_mgr->userq_count[queue->queue_type]); -- cgit v1.2.3 From a07930e4db4ceecb22f5ae63c97ed09c020844f8 Mon Sep 17 00:00:00 2001 From: Sunil Khatri Date: Fri, 6 Mar 2026 12:24:29 +0530 Subject: drm/amdgpu/userq: declutter the code with goto MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Clean up the amdgpu_userq_create function clean up in failure condition using goto method. This avoid replication of cleanup for every failure condition. Signed-off-by: Sunil Khatri Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 40 ++++++++++++++----------------- 1 file changed, 18 insertions(+), 22 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c index 049227fdb73a..4392a0fb6380 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c @@ -818,17 +818,15 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args) amdgpu_userq_input_va_validate(adev, queue, args->in.rptr_va, AMDGPU_GPU_PAGE_SIZE) || amdgpu_userq_input_va_validate(adev, queue, args->in.wptr_va, AMDGPU_GPU_PAGE_SIZE)) { r = -EINVAL; - kfree(queue); - goto unlock; + goto free_queue; } /* Convert relative doorbell offset into absolute doorbell index */ index = amdgpu_userq_get_doorbell_index(uq_mgr, &db_info, filp); if (index == (uint64_t)-EINVAL) { drm_file_err(uq_mgr->file, "Failed to get doorbell for queue\n"); - kfree(queue); r = -EINVAL; - goto unlock; + goto free_queue; } queue->doorbell_index = index; @@ -836,15 +834,13 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args) r = amdgpu_userq_fence_driver_alloc(adev, queue); if (r) { drm_file_err(uq_mgr->file, "Failed to alloc fence driver\n"); - goto unlock; + goto free_queue; } r = uq_funcs->mqd_create(queue, &args->in); if (r) { drm_file_err(uq_mgr->file, "Failed to create Queue\n"); - amdgpu_userq_fence_driver_free(queue); - kfree(queue); - goto unlock; + goto clean_fence_driver; } /* don't map the queue if scheduling is halted */ @@ -858,10 +854,8 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args) r = amdgpu_userq_map_helper(queue); if (r) { drm_file_err(uq_mgr->file, "Failed to map Queue\n"); - uq_funcs->mqd_destroy(queue); - amdgpu_userq_fence_driver_free(queue); - kfree(queue); - goto unlock; + down_read(&adev->reset_domain->sem); + goto clean_mqd; } } @@ -870,18 +864,15 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args) /* Wait for mode-1 reset to complete */ down_read(&adev->reset_domain->sem); + r = xa_alloc(&uq_mgr->userq_xa, &qid, queue, XA_LIMIT(1, AMDGPU_MAX_USERQ_COUNT), GFP_KERNEL); if (r) { if (!skip_map_queue) amdgpu_userq_unmap_helper(queue); - uq_funcs->mqd_destroy(queue); - amdgpu_userq_fence_driver_free(queue); - kfree(queue); r = -ENOMEM; - up_read(&adev->reset_domain->sem); - goto unlock; + goto clean_mqd; } r = xa_err(xa_store_irq(&adev->userq_doorbell_xa, index, queue, GFP_KERNEL)); @@ -890,11 +881,7 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args) if (!skip_map_queue) amdgpu_userq_unmap_helper(queue); - uq_funcs->mqd_destroy(queue); - amdgpu_userq_fence_driver_free(queue); - kfree(queue); - up_read(&adev->reset_domain->sem); - goto unlock; + goto clean_mqd; } up_read(&adev->reset_domain->sem); @@ -910,7 +897,16 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args) args->out.queue_id = qid; atomic_inc(&uq_mgr->userq_count[queue->queue_type]); + mutex_unlock(&uq_mgr->userq_mutex); + return 0; +clean_mqd: + uq_funcs->mqd_destroy(queue); + up_read(&adev->reset_domain->sem); +clean_fence_driver: + amdgpu_userq_fence_driver_free(queue); +free_queue: + kfree(queue); unlock: mutex_unlock(&uq_mgr->userq_mutex); -- cgit v1.2.3 From eb2e7f20c1473f9153088d564471dc0226530824 Mon Sep 17 00:00:00 2001 From: Sunil Khatri Date: Fri, 6 Mar 2026 16:45:14 +0530 Subject: drm/amdgpu: push userq debugfs function in amdgpu_debugfs files MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Debugfs files for amdgpu are better to be handled in the dedicated amdgpu_debugfs.c/.h files. Signed-off-by: Sunil Khatri Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 53 +++++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h | 4 +++ drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 49 +------------------------- 3 files changed, 58 insertions(+), 48 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c index b42f866935ab..aa9239b310a3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c @@ -39,6 +39,7 @@ #include "amdgpu_reset.h" #include "amdgpu_psp_ta.h" +#include "amdgpu_userq.h" #if defined(CONFIG_DEBUG_FS) @@ -2156,6 +2157,53 @@ static const struct file_operations amdgpu_pt_info_fops = { .release = single_release, }; +static int amdgpu_mqd_info_read(struct seq_file *m, void *unused) +{ + struct amdgpu_usermode_queue *queue = m->private; + struct amdgpu_bo *bo; + int r; + + if (!queue || !queue->mqd.obj) + return -EINVAL; + + bo = amdgpu_bo_ref(queue->mqd.obj); + r = amdgpu_bo_reserve(bo, true); + if (r) { + amdgpu_bo_unref(&bo); + return -EINVAL; + } + + seq_printf(m, "queue_type: %d\n", queue->queue_type); + seq_printf(m, "mqd_gpu_address: 0x%llx\n", amdgpu_bo_gpu_offset(queue->mqd.obj)); + + amdgpu_bo_unreserve(bo); + amdgpu_bo_unref(&bo); + + return 0; +} + +static int amdgpu_mqd_info_open(struct inode *inode, struct file *file) +{ + return single_open(file, amdgpu_mqd_info_read, inode->i_private); +} + +static const struct file_operations amdgpu_mqd_info_fops = { + .owner = THIS_MODULE, + .open = amdgpu_mqd_info_open, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release, +}; + +void amdgpu_debugfs_userq_init(struct drm_file *file, struct amdgpu_usermode_queue *queue, int qid) +{ + char queue_name[32]; + + scnprintf(queue_name, sizeof(queue_name), "queue_%d", qid); + queue->debugfs_queue = debugfs_create_dir(queue_name, file->debugfs_client); + debugfs_create_file("mqd_info", 0444, queue->debugfs_queue, queue, &amdgpu_mqd_info_fops); +} + void amdgpu_debugfs_vm_init(struct drm_file *file) { debugfs_create_file("vm_pagetable_info", 0444, file->debugfs_client, file, @@ -2174,4 +2222,9 @@ int amdgpu_debugfs_regs_init(struct amdgpu_device *adev) void amdgpu_debugfs_vm_init(struct drm_file *file) { } +void amdgpu_debugfs_userq_init(struct drm_file *file, + struct amdgpu_usermode_queue *queue, + int qid) +{ +} #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h index e7b3c38e5186..e88b4a1e564c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h @@ -25,6 +25,7 @@ /* * Debugfs */ +struct amdgpu_usermode_queue; int amdgpu_debugfs_regs_init(struct amdgpu_device *adev); int amdgpu_debugfs_init(struct amdgpu_device *adev); @@ -34,4 +35,7 @@ void amdgpu_debugfs_firmware_init(struct amdgpu_device *adev); void amdgpu_debugfs_gem_init(struct amdgpu_device *adev); void amdgpu_debugfs_mes_event_log_init(struct amdgpu_device *adev); void amdgpu_debugfs_vm_init(struct drm_file *file); +void amdgpu_debugfs_userq_init(struct drm_file *file, + struct amdgpu_usermode_queue *queue, + int qid); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c index 4392a0fb6380..7cdff206e7f3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c @@ -709,46 +709,6 @@ static int amdgpu_userq_priority_permit(struct drm_file *filp, return -EACCES; } -#if defined(CONFIG_DEBUG_FS) -static int amdgpu_mqd_info_read(struct seq_file *m, void *unused) -{ - struct amdgpu_usermode_queue *queue = m->private; - struct amdgpu_bo *bo; - int r; - - if (!queue || !queue->mqd.obj) - return -EINVAL; - - bo = amdgpu_bo_ref(queue->mqd.obj); - r = amdgpu_bo_reserve(bo, true); - if (r) { - amdgpu_bo_unref(&bo); - return -EINVAL; - } - - seq_printf(m, "queue_type: %d\n", queue->queue_type); - seq_printf(m, "mqd_gpu_address: 0x%llx\n", amdgpu_bo_gpu_offset(queue->mqd.obj)); - - amdgpu_bo_unreserve(bo); - amdgpu_bo_unref(&bo); - - return 0; -} - -static int amdgpu_mqd_info_open(struct inode *inode, struct file *file) -{ - return single_open(file, amdgpu_mqd_info_read, inode->i_private); -} - -static const struct file_operations amdgpu_mqd_info_fops = { - .owner = THIS_MODULE, - .open = amdgpu_mqd_info_open, - .read = seq_read, - .llseek = seq_lseek, - .release = single_release, -}; -#endif - static int amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args) { @@ -885,14 +845,7 @@ amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args) } up_read(&adev->reset_domain->sem); -#if defined(CONFIG_DEBUG_FS) - char queue_name[32]; - - scnprintf(queue_name, sizeof(queue_name), "queue_%d", qid); - /* Queue dentry per client to hold MQD information */ - queue->debugfs_queue = debugfs_create_dir(queue_name, filp->debugfs_client); - debugfs_create_file("mqd_info", 0444, queue->debugfs_queue, queue, &amdgpu_mqd_info_fops); -#endif + amdgpu_debugfs_userq_init(filp, queue, qid); amdgpu_userq_init_hang_detect_work(queue); args->out.queue_id = qid; -- cgit v1.2.3 From 6cafbc182855b8f456e04110b9111847446cd0ea Mon Sep 17 00:00:00 2001 From: Gaghik Khachatrian Date: Mon, 23 Feb 2026 14:56:15 -0500 Subject: drm/amd/display: Clean up NULL pointer warnings in dml2 This commit addresses multiple warnings by adding defensive checks for NULL pointers before dereferencing them. The changes ensure that pointers such as are validated, preventing potential undefined behavior. Reviewed-by: Dillon Varone Signed-off-by: Gaghik Khachatrian Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- .../drm/amd/display/dc/dml2_0/dml2_mall_phantom.c | 41 ++++++++++++++++++++-- 1 file changed, 38 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml2_mall_phantom.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml2_mall_phantom.c index 66040c877d68..d56e58ce26c7 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml2_mall_phantom.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml2_mall_phantom.c @@ -357,7 +357,7 @@ static bool enough_pipes_for_subvp(struct dml2_context *ctx, struct dc_state *st */ static bool subvp_subvp_schedulable(struct dml2_context *ctx, struct dc_state *context) { - struct pipe_ctx *subvp_pipes[2]; + struct pipe_ctx *subvp_pipes[2] = { NULL, NULL }; struct dc_stream_state *phantom = NULL; uint32_t microschedule_lines = 0; uint32_t index = 0; @@ -369,6 +369,9 @@ static bool subvp_subvp_schedulable(struct dml2_context *ctx, struct dc_state *c struct pipe_ctx *pipe = &context->res_ctx.pipe_ctx[i]; uint32_t time_us = 0; + if (pipe == NULL || pipe->stream == NULL) + continue; + /* Loop to calculate the maximum microschedule time between the two SubVP pipes, * and also to store the two main SubVP pipe pointers in subvp_pipes[2]. */ @@ -386,14 +389,19 @@ static bool subvp_subvp_schedulable(struct dml2_context *ctx, struct dc_state *c if (time_us > max_microschedule_us) max_microschedule_us = time_us; - subvp_pipes[index] = pipe; - index++; + if (index < 2) + subvp_pipes[index++] = pipe; // Maximum 2 SubVP pipes if (index == 2) break; } } + + /* Minimal guard to avoid C6001 before subvp_pipes[0]/[1] dereference */ + if (index < 2 || !subvp_pipes[0] || !subvp_pipes[1]) + return false; + vactive1_us = ((subvp_pipes[0]->stream->timing.v_addressable * subvp_pipes[0]->stream->timing.h_total) / (double)(subvp_pipes[0]->stream->timing.pix_clk_100hz * 100)) * 1000000; vactive2_us = ((subvp_pipes[1]->stream->timing.v_addressable * subvp_pipes[1]->stream->timing.h_total) / @@ -459,6 +467,11 @@ bool dml2_svp_drr_schedulable(struct dml2_context *ctx, struct dc_state *context break; } + if (pipe == NULL || pipe->stream == NULL) { + // Defensive: should never happen, try to catch in debug + ASSERT(0); + return false; + } phantom_stream = ctx->config.svp_pstate.callbacks.get_paired_subvp_stream(context, pipe->stream); main_timing = &pipe->stream->timing; phantom_timing = &phantom_stream->timing; @@ -549,6 +562,13 @@ static bool subvp_vblank_schedulable(struct dml2_context *ctx, struct dc_state * if (!subvp_pipe && pipe_mall_type == SUBVP_MAIN) subvp_pipe = pipe; } + + if (subvp_pipe == NULL) { + // Defensive: should never happen, catch in debug + ASSERT(0); + return false; + } + // Use ignore_msa_timing_param flag to identify as DRR if (found && context->res_ctx.pipe_ctx[vblank_index].stream->ignore_msa_timing_param) { // SUBVP + DRR case @@ -753,6 +773,12 @@ static void enable_phantom_plane(struct dml2_context *ctx, return; } + /* Minimal NULL guard for C6011 */ + if (!phantom_plane) { + ASSERT(0); + continue; + } + memcpy(&phantom_plane->address, &curr_pipe->plane_state->address, sizeof(phantom_plane->address)); memcpy(&phantom_plane->scaling_quality, &curr_pipe->plane_state->scaling_quality, sizeof(phantom_plane->scaling_quality)); @@ -880,6 +906,11 @@ bool dml2_svp_add_phantom_pipe_to_dc_state(struct dml2_context *ctx, struct dc_s if (ctx->config.svp_pstate.force_disable_subvp) return false; + if (!state) { + ASSERT(0); + return false; + } + if (!all_pipes_have_stream_and_plane(ctx, state)) return false; @@ -898,6 +929,10 @@ bool dml2_svp_add_phantom_pipe_to_dc_state(struct dml2_context *ctx, struct dc_s } if (enough_pipes_for_subvp(ctx, state) && assign_subvp_pipe(ctx, state, &dc_pipe_idx)) { + if (state->res_ctx.pipe_ctx[dc_pipe_idx].stream == NULL) { + ASSERT(0); + return false; + } dml_pipe_idx = dml2_helper_find_dml_pipe_idx_by_stream_id(ctx, state->res_ctx.pipe_ctx[dc_pipe_idx].stream->stream_id); svp_height = mode_support_info->SubViewportLinesNeededInMALL[dml_pipe_idx]; vstartup = dml_get_vstartup_calculated(&ctx->v20.dml_core_ctx, dml_pipe_idx); -- cgit v1.2.3 From aef320b01c4306957cac7be5468868806a27985c Mon Sep 17 00:00:00 2001 From: Peichen Huang Date: Fri, 6 Feb 2026 15:10:25 +0800 Subject: drm/amd/display: Return early from vesa replay enable function [WHY & HOW] If the enable state is already as expect then just return. Reviewed-by: Robin Chen Reviewed-by: Wenjing Liu Signed-off-by: Peichen Huang Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- .../dc/link/protocols/link_dp_panel_replay.c | 23 +++++++++++----------- 1 file changed, 12 insertions(+), 11 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_panel_replay.c b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_panel_replay.c index 6661078c0241..96afce4ffbfa 100644 --- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_panel_replay.c +++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_panel_replay.c @@ -252,23 +252,24 @@ bool dp_pr_enable(struct dc_link *link, bool enable) if (!dp_pr_get_panel_inst(dc, link, &panel_inst)) return false; + if (link->replay_settings.replay_allow_active == enable) + return true; + if (enable && !dc_is_embedded_signal(link->connector_signal)) dp_pr_set_static_screen_param(link); - if (link->replay_settings.replay_allow_active != enable) { - //for sending PR enable commands to DMUB - memset(&cmd, 0, sizeof(cmd)); + // for sending PR enable commands to DMUB + memset(&cmd, 0, sizeof(cmd)); - cmd.pr_enable.header.type = DMUB_CMD__PR; - cmd.pr_enable.header.sub_type = DMUB_CMD__PR_ENABLE; - cmd.pr_enable.header.payload_bytes = sizeof(struct dmub_cmd_pr_enable_data); - cmd.pr_enable.data.panel_inst = panel_inst; - cmd.pr_enable.data.enable = enable ? 1 : 0; + cmd.pr_enable.header.type = DMUB_CMD__PR; + cmd.pr_enable.header.sub_type = DMUB_CMD__PR_ENABLE; + cmd.pr_enable.header.payload_bytes = sizeof(struct dmub_cmd_pr_enable_data); + cmd.pr_enable.data.panel_inst = panel_inst; + cmd.pr_enable.data.enable = enable ? 1 : 0; + dc_wake_and_execute_dmub_cmd(dc->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT); - dc_wake_and_execute_dmub_cmd(dc->ctx, &cmd, DM_DMUB_WAIT_TYPE_WAIT); + link->replay_settings.replay_allow_active = enable; - link->replay_settings.replay_allow_active = enable; - } return true; } -- cgit v1.2.3 From 23dee18f6503d67b195f1513e404c78653ed0d40 Mon Sep 17 00:00:00 2001 From: Ovidiu Bunea Date: Tue, 10 Feb 2026 15:26:18 -0500 Subject: drm/amd/display: Add min clock init for DML21 mode programming [WHY & HOW] 0 stream cases do not go through any DML validation which leaves DCN clocks in unoptimized states. If requesting DML validation or programming with 0 streams, program DCN clocks to lowest DPM state. Reviewed-by: Dillon Varone Signed-off-by: Ovidiu Bunea Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- .../dc/dml2_0/dml21/dml21_translation_helper.c | 25 ++++++++++++++++++++++ .../dc/dml2_0/dml21/dml21_translation_helper.h | 1 + .../amd/display/dc/dml2_0/dml21/dml21_wrapper.c | 1 + 3 files changed, 27 insertions(+) diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_translation_helper.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_translation_helper.c index 75a279997961..cd12b7a6da65 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_translation_helper.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_translation_helper.c @@ -931,3 +931,28 @@ void dml21_set_dc_p_state_type( } } +void dml21_init_min_clocks_for_dc_state(struct dml2_context *in_ctx, struct dc_state *context) +{ + unsigned int lowest_dpm_state_index = 0; + struct dc_clocks *min_clocks = &context->bw_ctx.bw.dcn.clk; + + min_clocks->dispclk_khz = in_ctx->v21.dml_init.soc_bb.clk_table.dispclk.clk_values_khz[lowest_dpm_state_index]; + min_clocks->dppclk_khz = in_ctx->v21.dml_init.soc_bb.clk_table.dppclk.clk_values_khz[lowest_dpm_state_index]; + min_clocks->dcfclk_khz = in_ctx->v21.dml_init.soc_bb.clk_table.dcfclk.clk_values_khz[lowest_dpm_state_index]; + min_clocks->dramclk_khz = in_ctx->v21.dml_init.soc_bb.clk_table.uclk.clk_values_khz[lowest_dpm_state_index]; + min_clocks->fclk_khz = in_ctx->v21.dml_init.soc_bb.clk_table.fclk.clk_values_khz[lowest_dpm_state_index]; + min_clocks->idle_dramclk_khz = 0; + min_clocks->idle_fclk_khz = 0; + min_clocks->dcfclk_deep_sleep_khz = 0; + min_clocks->fclk_p_state_change_support = true; + min_clocks->p_state_change_support = true; + min_clocks->dtbclk_en = false; + min_clocks->ref_dtbclk_khz = 0; + min_clocks->socclk_khz = in_ctx->v21.dml_init.soc_bb.clk_table.socclk.clk_values_khz[lowest_dpm_state_index]; + min_clocks->subvp_prefetch_dramclk_khz = 0; + min_clocks->subvp_prefetch_fclk_khz = 0; + min_clocks->phyclk_khz = in_ctx->v21.dml_init.soc_bb.clk_table.phyclk.clk_values_khz[lowest_dpm_state_index]; + min_clocks->stutter_efficiency.base_efficiency = 1; + min_clocks->stutter_efficiency.low_power_efficiency = 1; +} + diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_translation_helper.h b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_translation_helper.h index 9880d3e0398e..f51d3d8a52c3 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_translation_helper.h +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_translation_helper.h @@ -25,4 +25,5 @@ void dml21_map_hw_resources(struct dml2_context *dml_ctx); void dml21_get_pipe_mcache_config(struct dc_state *context, struct pipe_ctx *pipe_ctx, struct dml2_per_plane_programming *pln_prog, struct dml2_pipe_configuration_descriptor *mcache_pipe_config); void dml21_set_dc_p_state_type(struct pipe_ctx *pipe_ctx, struct dml2_per_stream_programming *stream_programming, bool sub_vp_enabled); unsigned int map_plane_to_dml21_display_cfg(const struct dml2_context *dml_ctx, unsigned int stream_id, const struct dc_plane_state *plane, const struct dc_state *context); +void dml21_init_min_clocks_for_dc_state(struct dml2_context *in_ctx, struct dc_state *context); #endif diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_wrapper.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_wrapper.c index 798abb2b2e67..96c62bd6a37b 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_wrapper.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_wrapper.c @@ -215,6 +215,7 @@ static bool dml21_mode_check_and_programming(const struct dc *in_dc, struct dc_s return true; if (context->stream_count == 0) { + dml21_init_min_clocks_for_dc_state(dml_ctx, context); dml21_build_fams2_programming(in_dc, context, dml_ctx); return true; } -- cgit v1.2.3 From eb51c5a103f63b6e9c3f9cebf7b1e2e1056d3119 Mon Sep 17 00:00:00 2001 From: Harry Wentland Date: Fri, 16 Jan 2026 09:39:05 -0500 Subject: drm/amd/display: Set chroma taps to 1 if luma taps are 1 When luma is unscaled we also want chroma to be pixel-perfect. When luma taps are > 1 the result will be a blurred luma plane, even when the image isn't scaled. This makes IGT tests for CSC colorop pass. Assisted-by: Claude: claude-sonnet-4.5 Reviewed-by: Alex Hung Signed-off-by: Harry Wentland Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dpp/dcn10/dcn10_dpp.c | 8 ++++++-- drivers/gpu/drm/amd/display/dc/dpp/dcn30/dcn30_dpp.c | 8 ++++++-- 2 files changed, 12 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/dpp/dcn10/dcn10_dpp.c b/drivers/gpu/drm/amd/display/dc/dpp/dcn10/dcn10_dpp.c index ce91e5d28956..0e70ffc784b1 100644 --- a/drivers/gpu/drm/amd/display/dc/dpp/dcn10/dcn10_dpp.c +++ b/drivers/gpu/drm/amd/display/dc/dpp/dcn10/dcn10_dpp.c @@ -172,10 +172,14 @@ bool dpp1_get_optimal_number_of_taps( scl_data->taps.h_taps_c = in_taps->h_taps_c; if (!dpp->ctx->dc->debug.always_scale) { - if (IDENTITY_RATIO(scl_data->ratios.horz)) + if (IDENTITY_RATIO(scl_data->ratios.horz)) { scl_data->taps.h_taps = 1; - if (IDENTITY_RATIO(scl_data->ratios.vert)) + scl_data->taps.h_taps_c = 1; + } + if (IDENTITY_RATIO(scl_data->ratios.vert)) { scl_data->taps.v_taps = 1; + scl_data->taps.v_taps_c = 1; + } if (IDENTITY_RATIO(scl_data->ratios.horz_c)) scl_data->taps.h_taps_c = 1; if (IDENTITY_RATIO(scl_data->ratios.vert_c)) diff --git a/drivers/gpu/drm/amd/display/dc/dpp/dcn30/dcn30_dpp.c b/drivers/gpu/drm/amd/display/dc/dpp/dcn30/dcn30_dpp.c index c7923531da83..8a146968ee15 100644 --- a/drivers/gpu/drm/amd/display/dc/dpp/dcn30/dcn30_dpp.c +++ b/drivers/gpu/drm/amd/display/dc/dpp/dcn30/dcn30_dpp.c @@ -524,10 +524,14 @@ bool dpp3_get_optimal_number_of_taps( scl_data->taps.v_taps_c = max_taps_c; if (!dpp->ctx->dc->debug.always_scale) { - if (IDENTITY_RATIO(scl_data->ratios.horz)) + if (IDENTITY_RATIO(scl_data->ratios.horz)) { scl_data->taps.h_taps = 1; - if (IDENTITY_RATIO(scl_data->ratios.vert)) + scl_data->taps.h_taps_c = 1; + } + if (IDENTITY_RATIO(scl_data->ratios.vert)) { scl_data->taps.v_taps = 1; + scl_data->taps.v_taps_c = 1; + } if (IDENTITY_RATIO(scl_data->ratios.horz_c)) scl_data->taps.h_taps_c = 1; if (IDENTITY_RATIO(scl_data->ratios.vert_c)) -- cgit v1.2.3 From 63fff551318f5e0814b94f709a6dfaec789dcd7a Mon Sep 17 00:00:00 2001 From: Harry Wentland Date: Thu, 15 Jan 2026 14:12:05 -0500 Subject: drm/amd/display: Add NV12/P010 formats to primary plane Add NV12, NV21, and P010 YUV formats to the primary plane's supported format list, enabling YUV content to be scanned out directly from the primary plane. Assisted-by: Claude: claude-sonnet-4.5 Reviewed-by: Alex Hung Signed-off-by: Harry Wentland Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c index f8c21da62819..af069cce0cb2 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c @@ -62,6 +62,9 @@ static const uint32_t rgb_formats[] = { DRM_FORMAT_XBGR8888, DRM_FORMAT_ABGR8888, DRM_FORMAT_RGB565, + DRM_FORMAT_NV21, + DRM_FORMAT_NV12, + DRM_FORMAT_P010 }; static const uint32_t overlay_formats[] = { -- cgit v1.2.3 From f3d349484856e874ad00eeda9754a2ca8b7be5c7 Mon Sep 17 00:00:00 2001 From: Harry Wentland Date: Thu, 15 Jan 2026 14:12:46 -0500 Subject: drm/amd/display: Add COLOR_ENCODING/COLOR_RANGE to overlay planes Extend COLOR_ENCODING and COLOR_RANGE property creation to overlay planes in addition to primary planes. This allows overlay planes to use YUV formats with proper color space configuration when the hardware supports NV12/P010 formats. These properties control the YUV-to-RGB conversion matrix selection (BT.601/BT.709/BT.2020) and range handling (limited/full range). Assisted-by: Claude: claude-sonnet-4.5 Reviewed-by: Alex Hung Signed-off-by: Harry Wentland Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c index af069cce0cb2..d8c69fc94abb 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_plane.c @@ -1911,7 +1911,8 @@ int amdgpu_dm_plane_init(struct amdgpu_display_manager *dm, drm_plane_create_zpos_immutable_property(plane, 255); } - if (plane->type == DRM_PLANE_TYPE_PRIMARY && + if ((plane->type == DRM_PLANE_TYPE_PRIMARY || + plane->type == DRM_PLANE_TYPE_OVERLAY) && plane_cap && (plane_cap->pixel_format_support.nv12 || plane_cap->pixel_format_support.p010)) { -- cgit v1.2.3 From a879009f9a08b5756ca178223f003f18def07b22 Mon Sep 17 00:00:00 2001 From: Charlene Liu Date: Mon, 23 Feb 2026 14:28:14 -0500 Subject: drm/amd/display: Update underflow detection [WHY] Add underflow detection for later ASICs. Reviewed-by: Leo Chen Signed-off-by: Charlene Liu Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.h | 6 +++++- drivers/gpu/drm/amd/display/dc/optc/dcn31/dcn31_optc.c | 2 +- 2 files changed, 6 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.h b/drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.h index cf05620fd8f5..d2ba63046a50 100644 --- a/drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.h +++ b/drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.h @@ -218,7 +218,7 @@ uint32_t OTG_CRC_SIG_BLUE_CONTROL_MASK; \ uint32_t OTG_CRC_SIG_RED_GREEN_MASK; \ uint32_t OTG_DLPC_CONTROL; \ - uint32_t OTG_DRR_CONTROL2; \ + uint32_t OTG_DRR_CONTOL2; \ uint32_t OTG_DRR_TIMING_INT_STATUS; \ uint32_t OTG_GLOBAL_CONTROL3; \ uint32_t OTG_GLOBAL_SYNC_STATUS; \ @@ -676,6 +676,10 @@ struct dcn_optc_registers { type OTG_V_COUNT_STOP_TIMER; #define TG_REG_FIELD_LIST_DCN3_6(type) \ + type OPTC_RSMU_UNDERFLOW_CLEAR;\ + type OPTC_RSMU_UNDERFLOW_OCCURRED_STATUS;\ + type OPTC_RSMU_UNDERFLOW_INT_EN;\ + type OPTC_RSMU_UNDERFLOW_INT_STATUS;\ type OTG_CRC_POLY_SEL; \ type CRC0_R_CR32; \ type CRC0_G_Y32; \ diff --git a/drivers/gpu/drm/amd/display/dc/optc/dcn31/dcn31_optc.c b/drivers/gpu/drm/amd/display/dc/optc/dcn31/dcn31_optc.c index c6417538090f..893d2aff1f82 100644 --- a/drivers/gpu/drm/amd/display/dc/optc/dcn31/dcn31_optc.c +++ b/drivers/gpu/drm/amd/display/dc/optc/dcn31/dcn31_optc.c @@ -363,7 +363,7 @@ void optc31_read_reg_state(struct timing_generator *optc, struct dcn_optc_reg_st optc_reg_state->otg_crc3_data_rg = REG_READ(OTG_CRC3_DATA_RG); optc_reg_state->otg_dlpc_control = REG_READ(OTG_DLPC_CONTROL); optc_reg_state->otg_double_buffer_control = REG_READ(OTG_DOUBLE_BUFFER_CONTROL); - optc_reg_state->otg_drr_control2 = REG_READ(OTG_DRR_CONTROL2); + optc_reg_state->otg_drr_control2 = REG_READ(OTG_DRR_CONTOL2); optc_reg_state->otg_drr_control = REG_READ(OTG_DRR_CONTROL); optc_reg_state->otg_drr_timing_int_status = REG_READ(OTG_DRR_TIMING_INT_STATUS); optc_reg_state->otg_drr_trigger_window = REG_READ(OTG_DRR_TRIGGER_WINDOW); -- cgit v1.2.3 From 61436301dfc1d845f2b90a5f07d5c1d7cf89e900 Mon Sep 17 00:00:00 2001 From: Weiguang Li Date: Mon, 8 Dec 2025 14:13:20 +0800 Subject: drm/amd/display: Add ESD detection for replay recovery [HOW] Add Replay recovery flow so that when HPD occurs and ESD is detected, Replay can restore the system back to normal. Reviewed-by: Wenjing Liu Reviewed-by: Robin Chen Reviewed-by: Aric Cyr Signed-off-by: Weiguang Li Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- .../dc/link/protocols/link_dp_irq_handler.c | 36 +++++++++++++++------- drivers/gpu/drm/amd/display/include/dpcd_defs.h | 3 ++ 2 files changed, 28 insertions(+), 11 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_irq_handler.c b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_irq_handler.c index cc18a3bebef2..1860d44f63c1 100644 --- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_irq_handler.c +++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_irq_handler.c @@ -223,9 +223,10 @@ static void handle_hpd_irq_vesa_replay_sink(struct dc_link *link) } } -static void handle_hpd_irq_replay_sink(struct dc_link *link) +static void handle_hpd_irq_replay_sink(struct dc_link *link, bool *need_re_enable) { union dpcd_replay_configuration replay_configuration = {0}; + union dpcd_replay_configuration replay_sink_status = {0}; /*AMD Replay version reuse DP_PSR_ERROR_STATUS for REPLAY_ERROR status.*/ union psr_error_status replay_error_status = {0}; bool ret = false; @@ -265,9 +266,17 @@ static void handle_hpd_irq_replay_sink(struct dc_link *link) &replay_error_status.raw, sizeof(replay_error_status.raw)); + dm_helpers_dp_read_dpcd( + link->ctx, + link, + DP_PR_REPLAY_SINK_STATUS, + &replay_sink_status.raw, + 1); + if (replay_error_status.bits.LINK_CRC_ERROR || replay_configuration.bits.DESYNC_ERROR_STATUS || - replay_configuration.bits.STATE_TRANSITION_ERROR_STATUS) { + replay_configuration.bits.STATE_TRANSITION_ERROR_STATUS || + replay_sink_status.bits.SINK_DEVICE_REPLAY_STATUS == 0x7) { bool allow_active; link->replay_settings.config.replay_error_status.raw |= replay_error_status.raw; @@ -299,8 +308,7 @@ static void handle_hpd_irq_replay_sink(struct dc_link *link) if (link->replay_settings.replay_allow_active) { allow_active = false; edp_set_replay_allow_active(link, &allow_active, true, false, NULL); - allow_active = true; - edp_set_replay_allow_active(link, &allow_active, true, false, NULL); + *need_re_enable = true; } } } @@ -460,6 +468,7 @@ bool dp_handle_hpd_rx_irq(struct dc_link *link, union device_service_irq device_service_clear = {0}; enum dc_status result; bool status = false; + bool replay_re_enable_needed = false; if (out_link_loss) *out_link_loss = false; @@ -519,7 +528,7 @@ bool dp_handle_hpd_rx_irq(struct dc_link *link, /* PSR-related error was detected and handled */ return true; - handle_hpd_irq_replay_sink(link); + handle_hpd_irq_replay_sink(link, &replay_re_enable_needed); /* If PSR-related error handled, Main link may be off, * so do not handle as a normal sink status change interrupt. @@ -538,16 +547,16 @@ bool dp_handle_hpd_rx_irq(struct dc_link *link, return false; } - /* For now we only handle 'Downstream port status' case. + /* Handle 'Downstream port status' case for all DP link types. * If we got sink count changed it means * Downstream port status changed, * then DM should call DC to do the detection. - * NOTE: Do not handle link loss on eDP since it is internal link + * NOTE: Now includes eDP link loss detection and retraining */ - if ((link->connector_signal != SIGNAL_TYPE_EDP) && - dp_parse_link_loss_status( - link, - &hpd_irq_dpcd_data)) { + + if (dp_parse_link_loss_status( + link, + &hpd_irq_dpcd_data)) { /* Connectivity log: link loss */ CONN_DATA_LINK_LOSS(link, hpd_irq_dpcd_data.raw, @@ -576,6 +585,11 @@ bool dp_handle_hpd_rx_irq(struct dc_link *link, != link->dpcd_sink_count) status = true; + if (replay_re_enable_needed) { + bool allow_active = true; + + edp_set_replay_allow_active(link, &allow_active, true, false, NULL); + } /* reasons for HPD RX: * 1. Link Loss - ie Re-train the Link * 2. MST sideband message diff --git a/drivers/gpu/drm/amd/display/include/dpcd_defs.h b/drivers/gpu/drm/amd/display/include/dpcd_defs.h index 1afa10e85eb5..4a8ca0ac1266 100644 --- a/drivers/gpu/drm/amd/display/include/dpcd_defs.h +++ b/drivers/gpu/drm/amd/display/include/dpcd_defs.h @@ -64,6 +64,9 @@ #ifndef DP_PR_ERROR_STATUS // can remove this once the define gets into linux drm_dp_helper.h #define DP_PR_ERROR_STATUS 0x2020 /* DP 2.0 */ #endif /* DP_PR_ERROR_STATUS */ +#ifndef DP_PR_REPLAY_SINK_STATUS // can remove this once the define gets into linux drm_dp_helper.h +#define DP_PR_REPLAY_SINK_STATUS 0x2022 +#endif /* DP_PR_REPLAY_SINK_STATUS */ #ifndef DP_PR_LINK_CRC_ERROR // can remove this once the define gets into linux drm_dp_helper.h #define DP_PR_LINK_CRC_ERROR (1 << 0) #endif /* DP_PR_LINK_CRC_ERROR */ -- cgit v1.2.3 From e6e2b956fc814de766d3480be7018297c41d3ce0 Mon Sep 17 00:00:00 2001 From: Ivan Lipski Date: Tue, 24 Feb 2026 16:28:00 -0500 Subject: drm/amd/display: Add missing DCCG register entries for DCN20-DCN316 Commit 4c595e75110e ("drm/amd/display: Migrate DCCG registers access from hwseq to dccg component.") moved register writes from hwseq to dccg2_*() functions but did not add the registers to the DCCG register list macros. The struct fields default to 0, so REG_WRITE() targets MMIO offset 0, causing a GPU hang on resume (seen on DCN21/DCN30 during IGT kms_cursor_crc@cursor-suspend). Add - MICROSECOND_TIME_BASE_DIV - MILLISECOND_TIME_BASE_DIV - DCCG_GATE_DISABLE_CNTL - DCCG_GATE_DISABLE_CNTL2 - DC_MEM_GLOBAL_PWR_REQ_CNTL to macros in dcn20_dccg.h, dcn301_dccg.h, dcn31_dccg.h, and dcn314_dccg.h. Fixes: 4c595e75110e ("drm/amd/display: Migrate DCCG registers access from hwseq to dccg component.") Reported-by: Rafael Passos Reviewed-by: Aurabindo Pillai Signed-off-by: Ivan Lipski Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dccg/dcn20/dcn20_dccg.h | 6 +++++- drivers/gpu/drm/amd/display/dc/dccg/dcn301/dcn301_dccg.h | 8 +++++++- drivers/gpu/drm/amd/display/dc/dccg/dcn31/dcn31_dccg.h | 5 ++++- drivers/gpu/drm/amd/display/dc/dccg/dcn314/dcn314_dccg.h | 5 ++++- 4 files changed, 20 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/dccg/dcn20/dcn20_dccg.h b/drivers/gpu/drm/amd/display/dc/dccg/dcn20/dcn20_dccg.h index ffcd2e139e76..463f5826a1ea 100644 --- a/drivers/gpu/drm/amd/display/dc/dccg/dcn20/dcn20_dccg.h +++ b/drivers/gpu/drm/amd/display/dc/dccg/dcn20/dcn20_dccg.h @@ -38,7 +38,11 @@ DCCG_SRII(PIXEL_RATE_CNTL, OTG, 0),\ DCCG_SRII(PIXEL_RATE_CNTL, OTG, 1),\ SR(DISPCLK_FREQ_CHANGE_CNTL),\ - SR(DC_MEM_GLOBAL_PWR_REQ_CNTL) + SR(DC_MEM_GLOBAL_PWR_REQ_CNTL),\ + SR(MICROSECOND_TIME_BASE_DIV),\ + SR(MILLISECOND_TIME_BASE_DIV),\ + SR(DCCG_GATE_DISABLE_CNTL),\ + SR(DCCG_GATE_DISABLE_CNTL2) #define DCCG_REG_LIST_DCN2() \ DCCG_COMMON_REG_LIST_DCN_BASE(),\ diff --git a/drivers/gpu/drm/amd/display/dc/dccg/dcn301/dcn301_dccg.h b/drivers/gpu/drm/amd/display/dc/dccg/dcn301/dcn301_dccg.h index 067e49cb238e..e2381ca0be0b 100644 --- a/drivers/gpu/drm/amd/display/dc/dccg/dcn301/dcn301_dccg.h +++ b/drivers/gpu/drm/amd/display/dc/dccg/dcn301/dcn301_dccg.h @@ -34,7 +34,13 @@ DCCG_SRII(DTO_PARAM, DPPCLK, 1),\ DCCG_SRII(DTO_PARAM, DPPCLK, 2),\ DCCG_SRII(DTO_PARAM, DPPCLK, 3),\ - SR(REFCLK_CNTL) + SR(REFCLK_CNTL),\ + SR(DISPCLK_FREQ_CHANGE_CNTL),\ + SR(DC_MEM_GLOBAL_PWR_REQ_CNTL),\ + SR(MICROSECOND_TIME_BASE_DIV),\ + SR(MILLISECOND_TIME_BASE_DIV),\ + SR(DCCG_GATE_DISABLE_CNTL),\ + SR(DCCG_GATE_DISABLE_CNTL2) #define DCCG_MASK_SH_LIST_DCN301(mask_sh) \ DCCG_SFI(DPPCLK_DTO_CTRL, DTO_ENABLE, DPPCLK, 0, mask_sh),\ diff --git a/drivers/gpu/drm/amd/display/dc/dccg/dcn31/dcn31_dccg.h b/drivers/gpu/drm/amd/display/dc/dccg/dcn31/dcn31_dccg.h index bf659920d4cc..b5e3849ef12a 100644 --- a/drivers/gpu/drm/amd/display/dc/dccg/dcn31/dcn31_dccg.h +++ b/drivers/gpu/drm/amd/display/dc/dccg/dcn31/dcn31_dccg.h @@ -64,9 +64,12 @@ SR(DSCCLK1_DTO_PARAM),\ SR(DSCCLK2_DTO_PARAM),\ SR(DSCCLK_DTO_CTRL),\ + SR(DCCG_GATE_DISABLE_CNTL),\ SR(DCCG_GATE_DISABLE_CNTL2),\ SR(DCCG_GATE_DISABLE_CNTL3),\ - SR(HDMISTREAMCLK0_DTO_PARAM) + SR(HDMISTREAMCLK0_DTO_PARAM),\ + SR(DC_MEM_GLOBAL_PWR_REQ_CNTL),\ + SR(MICROSECOND_TIME_BASE_DIV) #define DCCG_MASK_SH_LIST_DCN31(mask_sh) \ diff --git a/drivers/gpu/drm/amd/display/dc/dccg/dcn314/dcn314_dccg.h b/drivers/gpu/drm/amd/display/dc/dccg/dcn314/dcn314_dccg.h index a609635f35db..ecbdc05f7c45 100644 --- a/drivers/gpu/drm/amd/display/dc/dccg/dcn314/dcn314_dccg.h +++ b/drivers/gpu/drm/amd/display/dc/dccg/dcn314/dcn314_dccg.h @@ -70,11 +70,14 @@ SR(DSCCLK2_DTO_PARAM),\ SR(DSCCLK3_DTO_PARAM),\ SR(DSCCLK_DTO_CTRL),\ + SR(DCCG_GATE_DISABLE_CNTL),\ SR(DCCG_GATE_DISABLE_CNTL2),\ SR(DCCG_GATE_DISABLE_CNTL3),\ SR(HDMISTREAMCLK0_DTO_PARAM),\ SR(OTG_PIXEL_RATE_DIV),\ - SR(DTBCLK_P_CNTL) + SR(DTBCLK_P_CNTL),\ + SR(DC_MEM_GLOBAL_PWR_REQ_CNTL),\ + SR(MICROSECOND_TIME_BASE_DIV) #define DCCG_MASK_SH_LIST_DCN314_COMMON(mask_sh) \ DCCG_SFI(DPPCLK_DTO_CTRL, DTO_DB_EN, DPPCLK, 0, mask_sh),\ -- cgit v1.2.3 From d3e79f333d90f6d3d268c3b073cf3afc0b019036 Mon Sep 17 00:00:00 2001 From: Nicholas Kazlauskas Date: Tue, 24 Feb 2026 16:49:46 -0500 Subject: drm/amd/display: Fix HWSS v3 fast path determination [WHY] We're checking surface and stream updates after they've been applied to their respective states within `update_planes_and_stream_state`. Medium updates under the HWSS V3 fast path that are not supported or tested are getting implicitly if they don't trigger a DML validation and getting updated in place on the dc->current_state context. [HOW] Fix this issue by moving up the fast path determination check prior to `update_planes_and_stream_state`. This is how the V2 path works and how the V3 path used to work prior to the refactors in this area. Reviewed-by: Dillon Varone Signed-off-by: Nicholas Kazlauskas Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/core/dc.c | 38 +++++++++++++++----------------- 1 file changed, 18 insertions(+), 20 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index 615bf2a01389..70f1c45ad834 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -7295,6 +7295,23 @@ static bool update_planes_and_stream_prepare_v3( ASSERT(scratch->flow == UPDATE_V3_FLOW_INVALID); dc_exit_ips_for_hw_access(scratch->dc); + /* HWSS path determination needs to be done prior to updating the surface and stream states. */ + struct dc_fast_update fast_update[MAX_SURFACES] = { 0 }; + + populate_fast_updates(fast_update, + scratch->surface_updates, + scratch->surface_count, + scratch->stream_update); + + const bool is_hwss_fast_path_only = + fast_update_only(scratch->dc, + fast_update, + scratch->surface_updates, + scratch->surface_count, + scratch->stream_update, + scratch->stream) && + !scratch->dc->check_config.enable_legacy_fast_update; + if (!update_planes_and_stream_state( scratch->dc, scratch->surface_updates, @@ -7310,26 +7327,7 @@ static bool update_planes_and_stream_prepare_v3( if (scratch->new_context == scratch->dc->current_state) { ASSERT(scratch->update_type < UPDATE_TYPE_FULL); - // TODO: Do we need this to be alive in execute? - struct dc_fast_update fast_update[MAX_SURFACES] = { 0 }; - - populate_fast_updates( - fast_update, - scratch->surface_updates, - scratch->surface_count, - scratch->stream_update - ); - const bool fast = fast_update_only( - scratch->dc, - fast_update, - scratch->surface_updates, - scratch->surface_count, - scratch->stream_update, - scratch->stream - ) - // TODO: Can this be used to skip `populate_fast_updates`? - && !scratch->dc->check_config.enable_legacy_fast_update; - scratch->flow = fast + scratch->flow = is_hwss_fast_path_only ? UPDATE_V3_FLOW_NO_NEW_CONTEXT_CONTEXT_FAST : UPDATE_V3_FLOW_NO_NEW_CONTEXT_CONTEXT_FULL; return true; -- cgit v1.2.3 From eff9280110bbea0b39586301a80a517cf5d79ce8 Mon Sep 17 00:00:00 2001 From: ChunTao Tso Date: Fri, 6 Feb 2026 15:41:54 +0800 Subject: drm/amd/display: Add new types to replay config [WHAT] Add FRAME_SKIPPING_ERROR_STATUS to dpcd_replay_configuration. Add received_frame_skipping_error_hpd to replay_config. Add REPLAY_GENERAL_CMD_SET_COASTING_VTOTAL_WITHOUT_FRAME_UPDATE to dmub_cmd_replay_general_subtype. Reviewed-by: Aric Cyr Signed-off-by: ChunTao Tso Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dc_dp_types.h | 2 +- drivers/gpu/drm/amd/display/dc/dc_types.h | 2 ++ drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h | 1 + 3 files changed, 4 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/display/dc/dc_dp_types.h b/drivers/gpu/drm/amd/display/dc/dc_dp_types.h index 9540f0ead279..7fa336bf1115 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_dp_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_dp_types.h @@ -1374,7 +1374,7 @@ union dpcd_replay_configuration { unsigned char DESYNC_ERROR_STATUS : 1; unsigned char SINK_DEVICE_REPLAY_STATUS : 3; unsigned char SINK_FRAME_LOCKED : 2; - unsigned char RESERVED : 1; + unsigned char FRAME_SKIPPING_ERROR_STATUS : 1; } bits; unsigned char raw; }; diff --git a/drivers/gpu/drm/amd/display/dc/dc_types.h b/drivers/gpu/drm/amd/display/dc/dc_types.h index d2e60480fb2b..d0ae5c27a511 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_types.h @@ -1194,6 +1194,8 @@ struct replay_config { union replay_optimization replay_optimization; /* Replay sub feature Frame Skipping is supported */ bool frame_skip_supported; + /* Replay Received Frame Skipping Error HPD. */ + bool received_frame_skipping_error_hpd; }; /* Replay feature flags*/ diff --git a/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h b/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h index 2abbc6c97850..f67a0ffb9335 100644 --- a/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h +++ b/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h @@ -4441,6 +4441,7 @@ enum dmub_cmd_replay_general_subtype { REPLAY_GENERAL_CMD_SET_LOW_RR_ACTIVATE, REPLAY_GENERAL_CMD_VIDEO_CONFERENCING, REPLAY_GENERAL_CMD_SET_CONTINUOUSLY_RESYNC, + REPLAY_GENERAL_CMD_SET_COASTING_VTOTAL_WITHOUT_FRAME_UPDATE, }; struct dmub_alpm_auxless_data { -- cgit v1.2.3 From b8469dd5b37fd25886c2018a04b71eb76fe8975d Mon Sep 17 00:00:00 2001 From: Nicholas Kazlauskas Date: Tue, 24 Feb 2026 10:36:29 -0500 Subject: drm/amd/display: Add documentation and cleanup DMUB HW lock manager To document how the helpers should be used before a subsequent change adds new locations that these helpers will be used. Reviewed-by: Ovidiu (Ovi) Bunea Signed-off-by: Nicholas Kazlauskas Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.h | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+) diff --git a/drivers/gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.h b/drivers/gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.h index 4c80ca8484ad..9495857902ed 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.h +++ b/drivers/gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.h @@ -46,7 +46,25 @@ void dmub_hw_lock_mgr_inbox0_cmd(struct dc_dmub_srv *dmub_srv, * Return: true if the inbox1 lock should be used, false otherwise */ bool should_use_dmub_inbox1_lock(const struct dc *dc, const struct dc_link *link); + +/** + * dmub_hw_lock_mgr_does_link_require_lock() - Returns true if the link has a feature that needs the HW lock. + * + * @dc: Pointer to DC object + * @link: The link to check + * + * Return: true if the link has a feature that needs the HW lock, false otherwise + */ bool dmub_hw_lock_mgr_does_link_require_lock(const struct dc *dc, const struct dc_link *link); + +/** + * dmub_hw_lock_mgr_does_context_require_lock() - Returns true if the context has any stream that needs the HW lock. + * + * @dc: Pointer to DC object + * @context: The context to check + * + * Return: true if the context has any stream that needs the HW lock, false otherwise + */ bool dmub_hw_lock_mgr_does_context_require_lock(const struct dc *dc, const struct dc_state *context); #endif /*_DMUB_HW_LOCK_MGR_H_ */ -- cgit v1.2.3 From c61eda434336cf2c033aa35efdc9a08b31d2fdfa Mon Sep 17 00:00:00 2001 From: Ivan Lipski Date: Wed, 25 Feb 2026 21:48:36 -0500 Subject: drm/amd/display: Check for S0i3 to be done before DCCG init on DCN21 [WHY] On DCN21, dccg2_init() is called in dcn10_init_hw() before bios_golden_init(). During S0i3 resume, BIOS sets MICROSECOND_TIME_BASE_DIV to 0x00120464 as a marker. dccg2_init() overwrites this to 0x00120264, causing dcn21_s0i3_golden_init_wa() to misdetect the state and skip golden init. Eventually during the resume sequence, a flip timeout occurs. [HOW] Skip DCCG on dccg2_is_s0i3_golden_init_wa_done() on DCN21. Fixes: 4c595e75110e ("drm/amd/display: Migrate DCCG registers access from hwseq to dccg component.") Reviewed-by: Aurabindo Pillai Signed-off-by: Ivan Lipski Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- .../gpu/drm/amd/display/dc/dccg/dcn21/dcn21_dccg.c | 21 ++++++++++++++++++++- 1 file changed, 20 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/display/dc/dccg/dcn21/dcn21_dccg.c b/drivers/gpu/drm/amd/display/dc/dccg/dcn21/dcn21_dccg.c index 75c69348027e..c4d4eea140f3 100644 --- a/drivers/gpu/drm/amd/display/dc/dccg/dcn21/dcn21_dccg.c +++ b/drivers/gpu/drm/amd/display/dc/dccg/dcn21/dcn21_dccg.c @@ -96,6 +96,25 @@ static void dccg21_update_dpp_dto(struct dccg *dccg, int dpp_inst, int req_dppcl dccg->pipe_dppclk_khz[dpp_inst] = req_dppclk; } +/* + * On DCN21 S0i3 resume, BIOS programs MICROSECOND_TIME_BASE_DIV to + * 0x00120464 as a marker that golden init has already been done. + * dcn21_s0i3_golden_init_wa() reads this marker later in bios_golden_init() + * to decide whether to skip golden init. + * + * dccg2_init() unconditionally overwrites MICROSECOND_TIME_BASE_DIV to + * 0x00120264, destroying the marker before it can be read. + * + * Guard the call: if the S0i3 marker is present, skip dccg2_init() so the + * WA can function correctly. bios_golden_init() will handle init in that case. + */ +static void dccg21_init(struct dccg *dccg) +{ + if (dccg2_is_s0i3_golden_init_wa_done(dccg)) + return; + + dccg2_init(dccg); +} static const struct dccg_funcs dccg21_funcs = { .update_dpp_dto = dccg21_update_dpp_dto, @@ -103,7 +122,7 @@ static const struct dccg_funcs dccg21_funcs = { .set_fifo_errdet_ovr_en = dccg2_set_fifo_errdet_ovr_en, .otg_add_pixel = dccg2_otg_add_pixel, .otg_drop_pixel = dccg2_otg_drop_pixel, - .dccg_init = dccg2_init, + .dccg_init = dccg21_init, .refclk_setup = dccg2_refclk_setup, /* Deprecated - for backward compatibility only */ .allow_clock_gating = dccg2_allow_clock_gating, .enable_memory_low_power = dccg2_enable_memory_low_power, -- cgit v1.2.3 From a28076422b0b1dc640471e4905d9c3a360b10a08 Mon Sep 17 00:00:00 2001 From: Gaghik Khachatrian Date: Thu, 26 Feb 2026 11:06:17 -0500 Subject: drm/amd/display: Fix compile warnings in dml2_0 Fix static analysis warnings by ensuring swath size temporaries are initialized before use. No functional change intended. Reviewed-by: Dillon Varone Signed-off-by: Gaghik Khachatrian Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dml2_0/display_mode_core.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/display_mode_core.c b/drivers/gpu/drm/amd/display/dc/dml2_0/display_mode_core.c index 09303c282495..8e8935995fca 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/display_mode_core.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/display_mode_core.c @@ -4089,8 +4089,8 @@ static void CalculateSwathAndDETConfiguration(struct display_mode_lib_scratch_st dml_uint_t MaximumSwathHeightC[__DML_NUM_PLANES__]; dml_uint_t RoundedUpMaxSwathSizeBytesY[__DML_NUM_PLANES__]; dml_uint_t RoundedUpMaxSwathSizeBytesC[__DML_NUM_PLANES__]; - dml_uint_t RoundedUpSwathSizeBytesY[__DML_NUM_PLANES__]; - dml_uint_t RoundedUpSwathSizeBytesC[__DML_NUM_PLANES__]; + dml_uint_t RoundedUpSwathSizeBytesY[__DML_NUM_PLANES__] = { 0 }; + dml_uint_t RoundedUpSwathSizeBytesC[__DML_NUM_PLANES__] = { 0 }; dml_uint_t SwathWidthSingleDPP[__DML_NUM_PLANES__]; dml_uint_t SwathWidthSingleDPPChroma[__DML_NUM_PLANES__]; -- cgit v1.2.3 From 41a5b6239fe8d437927d03f56816c7f167ac30dc Mon Sep 17 00:00:00 2001 From: Tom Chung Date: Thu, 26 Feb 2026 16:16:19 +0800 Subject: drm/amd/display: Add back missing memory type in array [WHY & HOW] Add back some missing memory type in window_memory_type. It should be aligned with enum dmub_window_id. Reviewed-by: Aurabindo Pillai Signed-off-by: Tom Chung Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c index 92f5c0b4d511..d650882a0a7a 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -2472,6 +2472,8 @@ static int dm_dmub_sw_init(struct amdgpu_device *adev) DMUB_WINDOW_MEMORY_TYPE_FB, //DMUB_WINDOW_7_SCRATCH_MEM DMUB_WINDOW_MEMORY_TYPE_FB, //DMUB_WINDOW_IB_MEM DMUB_WINDOW_MEMORY_TYPE_FB, //DMUB_WINDOW_SHARED_STATE + DMUB_WINDOW_MEMORY_TYPE_FB, //DMUB_WINDOW_LSDMA_BUFFER + DMUB_WINDOW_MEMORY_TYPE_FB, //DMUB_WINDOW_CURSOR_OFFLOAD }; int r; -- cgit v1.2.3 From 8c993dad415191e76169ca4b42eb00a8f8f7865a Mon Sep 17 00:00:00 2001 From: Clay King Date: Fri, 27 Feb 2026 12:34:34 -0500 Subject: drm/amd/display: Clean up unused code [WHAT] Silence warning by cleaning up unused code. Reviewed-by: Dillon Varone Signed-off-by: Clay King Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- .../amd/display/dc/clk_mgr/dcn30/dcn30_clk_mgr.c | 1 + drivers/gpu/drm/amd/display/dc/core/dc_resource.c | 1 - drivers/gpu/drm/amd/display/dc/dc_trace.h | 4 +-- .../amd/display/dc/hubbub/dcn401/dcn401_hubbub.c | 4 --- .../drm/amd/display/dc/hwss/dce110/dce110_hwseq.c | 24 ++++++++-------- .../drm/amd/display/dc/hwss/dcn10/dcn10_hwseq.c | 32 ++++++++-------------- .../drm/amd/display/dc/hwss/dcn31/dcn31_hwseq.c | 1 - .../drm/amd/display/dc/hwss/dcn35/dcn35_hwseq.c | 12 ++++---- .../drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c | 9 ------ .../gpu/drm/amd/display/dc/link/link_detection.c | 12 +------- .../dc/link/protocols/link_edp_panel_control.c | 2 -- .../gpu/drm/amd/display/dc/mpc/dcn10/dcn10_mpc.c | 4 +-- .../amd/display/dc/resource/dcn20/dcn20_resource.c | 2 -- 13 files changed, 35 insertions(+), 73 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30_clk_mgr.c index b0aba3a6f13c..b48522480dfd 100644 --- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30_clk_mgr.c +++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn30/dcn30_clk_mgr.c @@ -547,6 +547,7 @@ void dcn3_clk_mgr_construct( /* in case we don't get a value from the register, use default */ if (clk_mgr->base.dentist_vco_freq_khz == 0) clk_mgr->base.dentist_vco_freq_khz = 3650000; + /* Convert dprefclk units from MHz to KHz */ /* Value already divided by 10, some resolution lost */ diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_resource.c b/drivers/gpu/drm/amd/display/dc/core/dc_resource.c index c9fbb64d706a..8271b12c1a66 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_resource.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_resource.c @@ -2431,7 +2431,6 @@ static void resource_log_pipe_for_stream(struct dc *dc, struct dc_state *state, int slice_idx, dpp_idx, plane_idx, slice_count, dpp_count; bool is_primary; - DC_LOGGER_INIT(dc->ctx->logger); slice_count = resource_get_opp_heads_for_otg_master(otg_master, &state->res_ctx, opp_heads); diff --git a/drivers/gpu/drm/amd/display/dc/dc_trace.h b/drivers/gpu/drm/amd/display/dc/dc_trace.h index bbec308a3a5e..b7a011646d53 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_trace.h +++ b/drivers/gpu/drm/amd/display/dc/dc_trace.h @@ -23,8 +23,8 @@ #include "amdgpu_dm_trace.h" -#define TRACE_DC_PIPE_STATE(pipe_ctx, index, max_pipes) \ - for (index = 0; index < max_pipes; ++index) { \ +#define TRACE_DC_PIPE_STATE(pipe_ctx, max_pipes) \ + for (int index = 0; index < max_pipes; ++index) { \ struct pipe_ctx *pipe_ctx = &dc->current_state->res_ctx.pipe_ctx[index]; \ if (pipe_ctx->plane_state) \ trace_amdgpu_dm_dc_pipe_state(pipe_ctx->pipe_idx, pipe_ctx->plane_state, \ diff --git a/drivers/gpu/drm/amd/display/dc/hubbub/dcn401/dcn401_hubbub.c b/drivers/gpu/drm/amd/display/dc/hubbub/dcn401/dcn401_hubbub.c index 941dce439e97..b0a4b68cf359 100644 --- a/drivers/gpu/drm/amd/display/dc/hubbub/dcn401/dcn401_hubbub.c +++ b/drivers/gpu/drm/amd/display/dc/hubbub/dcn401/dcn401_hubbub.c @@ -1151,8 +1151,6 @@ void dcn401_program_compbuf_segments(struct hubbub *hubbub, unsigned compbuf_siz { struct dcn20_hubbub *hubbub2 = TO_DCN20_HUBBUB(hubbub); - unsigned int cur_compbuf_size_seg = 0; - if (safe_to_increase || compbuf_size_seg <= hubbub2->compbuf_size_segments) { if (compbuf_size_seg > hubbub2->compbuf_size_segments) { REG_WAIT(DCHUBBUB_DET0_CTRL, DET0_SIZE_CURRENT, hubbub2->det0_size, 1, 100); @@ -1165,8 +1163,6 @@ void dcn401_program_compbuf_segments(struct hubbub *hubbub, unsigned compbuf_siz + hubbub2->det3_size + compbuf_size_seg <= hubbub2->crb_size_segs); REG_UPDATE(DCHUBBUB_COMPBUF_CTRL, COMPBUF_SIZE, compbuf_size_seg); hubbub2->compbuf_size_segments = compbuf_size_seg; - - ASSERT(REG_GET(DCHUBBUB_COMPBUF_CTRL, CONFIG_ERROR, &cur_compbuf_size_seg) && !cur_compbuf_size_seg); } } diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c index 8aafd460c36f..38a5f312ddb4 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c @@ -86,9 +86,9 @@ hws->ctx #define DC_LOGGER \ - ctx->logger -#define DC_LOGGER_INIT() \ - struct dc_context *ctx = dc->ctx + dc_ctx->logger +#define DC_LOGGER_INIT(ctx) \ + struct dc_context *dc_ctx = ctx #define REG(reg)\ hws->regs->reg @@ -687,7 +687,7 @@ dce110_external_encoder_control(enum bp_external_encoder_control_action action, .pixel_clock = timing ? timing->pix_clk_100hz / 10 : 300000, .color_depth = timing ? timing->display_color_depth : COLOR_DEPTH_888, }; - DC_LOGGER_INIT(); + DC_LOGGER_INIT(dc->ctx); bp_result = bios->funcs->external_encoder_control(bios, &ext_cntl); @@ -767,13 +767,14 @@ void dce110_edp_wait_for_hpd_ready( struct dc_link *link, bool power_up) { - struct dc_context *ctx = link->ctx; struct graphics_object_id connector = link->link_enc->connector; bool edp_hpd_high = false; uint32_t time_elapsed = 0; uint32_t timeout = power_up ? PANEL_POWER_UP_TIMEOUT : PANEL_POWER_DOWN_TIMEOUT; + DC_LOGGER_INIT(link->ctx); + if (dal_graphics_object_id_get_connector_id(connector) != CONNECTOR_ID_EDP) { BREAK_TO_DEBUGGER(); @@ -825,6 +826,7 @@ void dce110_edp_power_control( enum bp_result bp_result; uint8_t pwrseq_instance; + DC_LOGGER_INIT(ctx); if (dal_graphics_object_id_get_connector_id(link->link_enc->connector) != CONNECTOR_ID_EDP) { @@ -993,6 +995,8 @@ void dce110_edp_backlight_control( unsigned int pre_T11_delay = (link->dpcd_sink_ext_caps.bits.oled ? OLED_PRE_T11_DELAY : 0); unsigned int post_T7_delay = (link->dpcd_sink_ext_caps.bits.oled ? OLED_POST_T7_DELAY : 0); + DC_LOGGER_INIT(ctx); + if (dal_graphics_object_id_get_connector_id(link->link_enc->connector) != CONNECTOR_ID_EDP) { BREAK_TO_DEBUGGER(); @@ -1969,8 +1973,7 @@ void dce110_enable_accelerated_mode(struct dc *dc, struct dc_state *context) bool keep_edp_vdd_on = false; bool should_clean_dsc_block = true; struct dc_bios *dcb = dc->ctx->dc_bios; - DC_LOGGER_INIT(); - + DC_LOGGER_INIT(dc->ctx); get_edp_links_with_sink(dc, edp_links_with_sink, &edp_with_sink_num); dc_get_edp_links(dc, edp_links, &edp_num); @@ -2736,7 +2739,6 @@ static bool wait_for_reset_trigger_to_occur( struct dc_context *dc_ctx, struct timing_generator *tg) { - struct dc_context *ctx = dc_ctx; bool rc = false; /* To avoid endless loop we wait at most @@ -2778,10 +2780,9 @@ static void dce110_enable_timing_synchronization( int group_size, struct pipe_ctx *grouped_pipes[]) { - struct dc_context *dc_ctx = dc->ctx; struct dcp_gsl_params gsl_params = { 0 }; int i; - DC_LOGGER_INIT(); + DC_LOGGER_INIT(dc->ctx); DC_SYNC_INFO("GSL: Setting-up...\n"); @@ -2824,10 +2825,9 @@ static void dce110_enable_per_frame_crtc_position_reset( int group_size, struct pipe_ctx *grouped_pipes[]) { - struct dc_context *dc_ctx = dc->ctx; struct dcp_gsl_params gsl_params = { 0 }; int i; - DC_LOGGER_INIT(); + DC_LOGGER_INIT(dc->ctx); gsl_params.gsl_group = 0; gsl_params.gsl_master = 0; diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn10/dcn10_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn10/dcn10_hwseq.c index a2d28be480e8..17ff66d9a617 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn10/dcn10_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn10/dcn10_hwseq.c @@ -60,9 +60,9 @@ #include "dc_state_priv.h" #define DC_LOGGER \ - dc_logger -#define DC_LOGGER_INIT(logger) \ - struct dal_logger *dc_logger = logger + dc_ctx->logger +#define DC_LOGGER_INIT(ctx) \ + struct dc_context *dc_ctx = ctx #define CTX \ hws->ctx @@ -1009,7 +1009,7 @@ static void power_on_plane_resources( struct dce_hwseq *hws, int plane_id) { - DC_LOGGER_INIT(hws->ctx->logger); + DC_LOGGER_INIT(hws->ctx); if (hws->funcs.dpp_root_clock_control) hws->funcs.dpp_root_clock_control(hws, plane_id, true); @@ -1286,7 +1286,7 @@ static void dcn10_reset_back_end_for_pipe( { int i; struct dc_link *link; - DC_LOGGER_INIT(dc->ctx->logger); + DC_LOGGER_INIT(dc->ctx); if (pipe_ctx->stream_res.stream_enc == NULL) { pipe_ctx->stream = NULL; return; @@ -1422,12 +1422,10 @@ void dcn10_verify_allow_pstate_change_high(struct dc *dc) return; if (!hubbub->funcs->verify_allow_pstate_change_high(hubbub)) { - int i = 0; - if (should_log_hw_state) dcn10_log_hw_state(dc, NULL); - TRACE_DC_PIPE_STATE(pipe_ctx, i, MAX_PIPES); + TRACE_DC_PIPE_STATE(pipe_ctx, MAX_PIPES); BREAK_TO_DEBUGGER(); if (dcn10_hw_wa_force_recovery(dc)) { /*check again*/ @@ -1490,7 +1488,7 @@ void dcn10_plane_atomic_power_down(struct dc *dc, struct hubp *hubp) { struct dce_hwseq *hws = dc->hwseq; - DC_LOGGER_INIT(dc->ctx->logger); + DC_LOGGER_INIT(dc->ctx); if (REG(DC_IP_REQUEST_CNTL)) { REG_SET(DC_IP_REQUEST_CNTL, 0, @@ -1554,7 +1552,7 @@ void dcn10_plane_atomic_disable(struct dc *dc, struct pipe_ctx *pipe_ctx) void dcn10_disable_plane(struct dc *dc, struct dc_state *state, struct pipe_ctx *pipe_ctx) { struct dce_hwseq *hws = dc->hwseq; - DC_LOGGER_INIT(dc->ctx->logger); + DC_LOGGER_INIT(dc->ctx); if (!pipe_ctx->plane_res.hubp || pipe_ctx->plane_res.hubp->power_gated) return; @@ -2268,8 +2266,6 @@ static bool wait_for_reset_trigger_to_occur( { bool rc = false; - DC_LOGGER_INIT(dc_ctx->logger); - /* To avoid endless loop we wait at most * frames_to_wait_on_triggered_reset frames for the reset to occur. */ const uint32_t frames_to_wait_on_triggered_reset = 10; @@ -2384,7 +2380,6 @@ static uint8_t get_clock_divider(struct pipe_ctx *pipe, static int dcn10_align_pixel_clocks(struct dc *dc, int group_size, struct pipe_ctx *grouped_pipes[]) { - struct dc_context *dc_ctx = dc->ctx; int i, master = -1, embedded = -1; struct dc_crtc_timing *hw_crtc_timing; uint64_t phase[MAX_PIPES]; @@ -2397,7 +2392,7 @@ static int dcn10_align_pixel_clocks(struct dc *dc, int group_size, uint32_t dp_ref_clk_100hz = dc->res_pool->dp_clock_source->ctx->dc->clk_mgr->dprefclk_khz*10; - DC_LOGGER_INIT(dc_ctx->logger); + DC_LOGGER_INIT(dc->ctx); hw_crtc_timing = kzalloc_objs(*hw_crtc_timing, MAX_PIPES); if (!hw_crtc_timing) @@ -2477,12 +2472,11 @@ void dcn10_enable_vblanks_synchronization( int group_size, struct pipe_ctx *grouped_pipes[]) { - struct dc_context *dc_ctx = dc->ctx; struct output_pixel_processor *opp; struct timing_generator *tg; int i, width = 0, height = 0, master; - DC_LOGGER_INIT(dc_ctx->logger); + DC_LOGGER_INIT(dc->ctx); for (i = 1; i < group_size; i++) { opp = grouped_pipes[i]->stream_res.opp; @@ -2543,12 +2537,11 @@ void dcn10_enable_timing_synchronization( int group_size, struct pipe_ctx *grouped_pipes[]) { - struct dc_context *dc_ctx = dc->ctx; struct output_pixel_processor *opp; struct timing_generator *tg; int i, width = 0, height = 0; - DC_LOGGER_INIT(dc_ctx->logger); + DC_LOGGER_INIT(dc->ctx); DC_SYNC_INFO("Setting up OTG reset trigger\n"); @@ -2624,10 +2617,9 @@ void dcn10_enable_per_frame_crtc_position_reset( int group_size, struct pipe_ctx *grouped_pipes[]) { - struct dc_context *dc_ctx = dc->ctx; int i; - DC_LOGGER_INIT(dc_ctx->logger); + DC_LOGGER_INIT(dc->ctx); DC_SYNC_INFO("Setting up\n"); for (i = 0; i < group_size; i++) diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn31/dcn31_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn31/dcn31_hwseq.c index db2f7cbb12ff..94f63fd54e3e 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn31/dcn31_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn31/dcn31_hwseq.c @@ -513,7 +513,6 @@ static void dcn31_reset_back_end_for_pipe( { struct dc_link *link; - DC_LOGGER_INIT(dc->ctx->logger); if (pipe_ctx->stream_res.stream_enc == NULL) { pipe_ctx->stream = NULL; return; diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn35/dcn35_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn35/dcn35_hwseq.c index b5a4cefbd35f..b5f60f59382e 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn35/dcn35_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn35/dcn35_hwseq.c @@ -60,15 +60,15 @@ #include "dcn20/dcn20_hwseq.h" #include "dc_state_priv.h" -#define DC_LOGGER_INIT(logger) \ - struct dal_logger *dc_logger = logger +#define DC_LOGGER \ + dc_ctx->logger +#define DC_LOGGER_INIT(ctx) \ + struct dc_context *dc_ctx = ctx #define CTX \ hws->ctx #define REG(reg)\ hws->regs->reg -#define DC_LOGGER \ - dc_logger #undef FN @@ -331,7 +331,7 @@ static void update_dsc_on_stream(struct pipe_ctx *pipe_ctx, bool enable) struct pipe_ctx *odm_pipe; int opp_cnt = 1; - DC_LOGGER_INIT(stream->ctx->logger); + DC_LOGGER_INIT(stream->ctx); ASSERT(dsc); for (odm_pipe = pipe_ctx->next_odm_pipe; odm_pipe; odm_pipe = odm_pipe->next_odm_pipe) @@ -897,7 +897,7 @@ void dcn35_disable_plane(struct dc *dc, struct dc_state *state, struct pipe_ctx bool is_phantom = dc_state_get_pipe_subvp_type(state, pipe_ctx) == SUBVP_PHANTOM; struct timing_generator *tg = is_phantom ? pipe_ctx->stream_res.tg : NULL; - DC_LOGGER_INIT(dc->ctx->logger); + DC_LOGGER_INIT(dc->ctx); if (!pipe_ctx->plane_res.hubp || pipe_ctx->plane_res.hubp->power_gated) return; diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c index 69cc70106bf0..55c48c4bc6ce 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c @@ -1982,7 +1982,6 @@ void dcn401_reset_back_end_for_pipe( struct dc_link *link = pipe_ctx->stream->link; const struct link_hwss *link_hwss = get_link_hwss(link, &pipe_ctx->link_res); - DC_LOGGER_INIT(dc->ctx->logger); if (pipe_ctx->stream_res.stream_enc == NULL) { pipe_ctx->stream = NULL; return; @@ -2422,8 +2421,6 @@ void dcn401_program_front_end_for_ctx( struct dce_hwseq *hws = dc->hwseq; struct pipe_ctx *pipe = NULL; - DC_LOGGER_INIT(dc->ctx->logger); - if (resource_is_pipe_topology_changed(dc->current_state, context)) resource_log_pipe_topology_update(dc, context); @@ -2587,8 +2584,6 @@ void dcn401_post_unlock_program_front_end( struct dce_hwseq *hwseq = dc->hwseq; int i; - DC_LOGGER_INIT(dc->ctx->logger); - for (i = 0; i < dc->res_pool->pipe_count; i++) if (resource_is_pipe_type(&dc->current_state->res_ctx.pipe_ctx[i], OPP_HEAD) && !resource_is_pipe_type(&context->res_ctx.pipe_ctx[i], OPP_HEAD)) @@ -2968,8 +2963,6 @@ void dcn401_plane_atomic_power_down(struct dc *dc, struct dce_hwseq *hws = dc->hwseq; uint32_t org_ip_request_cntl = 0; - DC_LOGGER_INIT(dc->ctx->logger); - if (REG(DC_IP_REQUEST_CNTL)) { REG_GET(DC_IP_REQUEST_CNTL, IP_REQUEST_EN, &org_ip_request_cntl); if (org_ip_request_cntl == 0) @@ -3061,8 +3054,6 @@ void dcn401_plane_atomic_power_down_sequence(struct dc *dc, struct dce_hwseq *hws = dc->hwseq; uint32_t org_ip_request_cntl = 0; - DC_LOGGER_INIT(dc->ctx->logger); - /* Check and set DC_IP_REQUEST_CNTL if needed */ if (REG(DC_IP_REQUEST_CNTL)) { REG_GET(DC_IP_REQUEST_CNTL, IP_REQUEST_EN, &org_ip_request_cntl); diff --git a/drivers/gpu/drm/amd/display/dc/link/link_detection.c b/drivers/gpu/drm/amd/display/dc/link/link_detection.c index d0bb26888f4b..f992c2d16748 100644 --- a/drivers/gpu/drm/amd/display/dc/link/link_detection.c +++ b/drivers/gpu/drm/amd/display/dc/link/link_detection.c @@ -352,7 +352,7 @@ static void query_dp_dual_mode_adaptor( *dongle = DISPLAY_DONGLE_DP_DVI_DONGLE; sink_cap->max_hdmi_pixel_clock = DP_ADAPTOR_DVI_MAX_TMDS_CLK; - CONN_DATA_DETECT(ddc->link, type2_dongle_buf, sizeof(type2_dongle_buf), + CONN_DATA_DETECT(link, type2_dongle_buf, sizeof(type2_dongle_buf), "DP-DVI passive dongle %dMhz: ", DP_ADAPTOR_DVI_MAX_TMDS_CLK / 1000); return; @@ -657,8 +657,6 @@ static bool wait_for_entering_dp_alt_mode(struct dc_link *link) unsigned long long time_taken_in_ns; int tries_taken; - DC_LOGGER_INIT(link->ctx->logger); - /** * this function will only exist if we are on dcn21 (is_in_alt_mode is a * function pointer, so checking to see if it is equal to 0 is the same @@ -729,8 +727,6 @@ static void revert_dpia_mst_dsc_always_on_wa(struct dc_link *link) static bool discover_dp_mst_topology(struct dc_link *link, enum dc_detect_reason reason) { - DC_LOGGER_INIT(link->ctx->logger); - LINK_INFO("link=%d, mst branch is now Connected\n", link->link_index); @@ -750,8 +746,6 @@ static bool discover_dp_mst_topology(struct dc_link *link, enum dc_detect_reason bool link_reset_cur_dp_mst_topology(struct dc_link *link) { - DC_LOGGER_INIT(link->ctx->logger); - LINK_INFO("link=%d, mst branch is now Disconnected\n", link->link_index); @@ -977,8 +971,6 @@ static bool detect_link_and_local_sink(struct dc_link *link, enum dc_connection_type new_connection_type = dc_connection_none; const uint32_t post_oui_delay = 30; // 30ms - DC_LOGGER_INIT(link->ctx->logger); - if (dc_is_virtual_signal(link->connector_signal)) return false; @@ -1459,8 +1451,6 @@ bool link_detect(struct dc_link *link, enum dc_detect_reason reason) bool is_delegated_to_mst_top_mgr = false; enum dc_connection_type pre_link_type = link->type; - DC_LOGGER_INIT(link->ctx->logger); - is_local_sink_detect_success = detect_link_and_local_sink(link, reason); if (is_local_sink_detect_success && link->local_sink) diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_edp_panel_control.c b/drivers/gpu/drm/amd/display/dc/link/protocols/link_edp_panel_control.c index 5b2c1a4911cf..ba311a8f3f65 100644 --- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_edp_panel_control.c +++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_edp_panel_control.c @@ -1094,8 +1094,6 @@ bool edp_send_replay_cmd(struct dc_link *link, if (!replay) return false; - DC_LOGGER_INIT(link->ctx->logger); - if (dp_pr_get_panel_inst(dc, link, &panel_inst)) cmd_data->panel_inst = panel_inst; else { diff --git a/drivers/gpu/drm/amd/display/dc/mpc/dcn10/dcn10_mpc.c b/drivers/gpu/drm/amd/display/dc/mpc/dcn10/dcn10_mpc.c index b23c64004dd5..27e653234850 100644 --- a/drivers/gpu/drm/amd/display/dc/mpc/dcn10/dcn10_mpc.c +++ b/drivers/gpu/drm/amd/display/dc/mpc/dcn10/dcn10_mpc.c @@ -118,9 +118,7 @@ void mpc1_assert_idle_mpcc(struct mpc *mpc, int id) struct mpcc *mpc1_get_mpcc(struct mpc *mpc, int mpcc_id) { - struct dcn10_mpc *mpc10 = TO_DCN10_MPC(mpc); - - ASSERT(mpcc_id < mpc10->num_mpcc); + ASSERT(mpcc_id < TO_DCN10_MPC(mpc)->num_mpcc); return &(mpc->mpcc_array[mpcc_id]); } diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn20/dcn20_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn20/dcn20_resource.c index 8b555187ac75..366576b1c617 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn20/dcn20_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn20/dcn20_resource.c @@ -2341,8 +2341,6 @@ static bool init_soc_bounding_box(struct dc *dc, struct _vcs_dpi_ip_params_st *loaded_ip = get_asic_rev_ip_params(dc->ctx->asic_id.hw_internal_rev); - DC_LOGGER_INIT(dc->ctx->logger); - if (pool->base.pp_smu) { struct pp_smu_nv_clock_table max_clocks = {0}; unsigned int uclk_states[8] = {0}; -- cgit v1.2.3 From 5b880e37a89feda5185cd1875b9d1d213dc731d1 Mon Sep 17 00:00:00 2001 From: Roman Li Date: Fri, 20 Feb 2026 16:48:14 -0500 Subject: drm/amd/display: Enable dcn42 DC clk_mgr Add support for DCN 4.2 clock manager. Acked-by: Aurabindo Pillai Signed-off-by: Roman Li Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/clk_mgr/clk_mgr.c | 16 + .../amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c | 236 +++++++------- drivers/gpu/drm/amd/display/dc/core/dc.c | 1 + drivers/gpu/drm/amd/display/dc/dc_spl_translate.c | 1 + .../gpu/drm/amd/display/dc/dccg/dcn20/dcn20_dccg.h | 3 +- .../gpu/drm/amd/display/dc/dccg/dcn42/dcn42_dccg.c | 59 +++- .../gpu/drm/amd/display/dc/dccg/dcn42/dcn42_dccg.h | 11 +- .../dml2_0/dml21/inc/bounding_boxes/dcn42_soc_bb.h | 359 +++++++++++++++++++++ drivers/gpu/drm/amd/display/dc/inc/hw/clk_mgr.h | 2 +- .../drm/amd/display/dc/inc/hw/clk_mgr_internal.h | 73 ++++- .../dc/link/protocols/link_edp_panel_control.c | 1 + .../amd/display/dc/soc_and_ip_translator/Makefile | 3 + .../dcn42/dcn42_soc_and_ip_translator.c | 159 ++++++++- .../dcn42/dcn42_soc_and_ip_translator.h | 1 + .../soc_and_ip_translator/soc_and_ip_translator.c | 4 + .../amd/include/asic_reg/clk/clk_15_0_0_offset.h | 44 +++ .../amd/include/asic_reg/clk/clk_15_0_0_sh_mask.h | 52 +++ 17 files changed, 885 insertions(+), 140 deletions(-) create mode 100644 drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/bounding_boxes/dcn42_soc_bb.h create mode 100644 drivers/gpu/drm/amd/include/asic_reg/clk/clk_15_0_0_offset.h create mode 100644 drivers/gpu/drm/amd/include/asic_reg/clk/clk_15_0_0_sh_mask.h diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/clk_mgr.c index 5ad66d873aad..e91636d033c7 100644 --- a/drivers/gpu/drm/amd/display/dc/clk_mgr/clk_mgr.c +++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/clk_mgr.c @@ -48,6 +48,7 @@ #include "dcn32/dcn32_clk_mgr.h" #include "dcn35/dcn35_clk_mgr.h" #include "dcn401/dcn401_clk_mgr.h" +#include "dcn42/dcn42_clk_mgr.h" int clk_mgr_helper_get_active_display_cnt( struct dc *dc, @@ -362,6 +363,18 @@ struct clk_mgr *dc_clk_mgr_create(struct dc_context *ctx, struct pp_smu_funcs *p return &clk_mgr->base; } break; + case AMDGPU_FAMILY_GC_11_5_4: { + struct clk_mgr_dcn42 *clk_mgr = kzalloc(sizeof(*clk_mgr), GFP_KERNEL); + + if (clk_mgr == NULL) { + BREAK_TO_DEBUGGER(); + return NULL; + } + + dcn42_clk_mgr_construct(ctx, clk_mgr, pp_smu, dccg); + return &clk_mgr->base.base; + } + break; #endif /* CONFIG_DRM_AMD_DC_FP */ default: ASSERT(0); /* Unknown Asic */ @@ -419,6 +432,9 @@ void dc_destroy_clk_mgr(struct clk_mgr *clk_mgr_base) case AMDGPU_FAMILY_GC_12_0_0: dcn401_clk_mgr_destroy(clk_mgr); break; + case AMDGPU_FAMILY_GC_11_5_4: + dcn42_clk_mgr_destroy(clk_mgr); + break; default: break; diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c index 55434f046fa2..cbb79789a532 100644 --- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c +++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c @@ -31,59 +31,19 @@ #include "link_service.h" #include "logger_types.h" +#include "clk/clk_15_0_0_offset.h" +#include "clk/clk_15_0_0_sh_mask.h" +#include "dcn/dcn_4_2_0_offset.h" +#include "dcn/dcn_4_2_0_sh_mask.h" + #undef DC_LOGGER #define DC_LOGGER \ - clk_mgr->base.base.ctx->logger - - -#define DCN_BASE__INST0_SEG1 0x000000C0 - -#define regCLK8_CLK2_BYPASS_CNTL 0x4c2a -#define regCLK8_CLK2_BYPASS_CNTL_BASE_IDX 0 -#define CLK8_CLK2_BYPASS_CNTL__CLK2_BYPASS_SEL__SHIFT 0x0 -#define CLK8_CLK2_BYPASS_CNTL__CLK2_BYPASS_DIV__SHIFT 0x10 -#define CLK8_CLK2_BYPASS_CNTL__CLK2_BYPASS_SEL_MASK 0x00000007L -#define CLK8_CLK2_BYPASS_CNTL__CLK2_BYPASS_DIV_MASK 0x000F0000L - -#define regDENTIST_DISPCLK_CNTL 0x0064 -#define regDENTIST_DISPCLK_CNTL_BASE_IDX 1 - -// DENTIST_DISPCLK_CNTL -#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_WDIVIDER__SHIFT 0x0 -#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_RDIVIDER__SHIFT 0x8 -#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_CHG_DONE__SHIFT 0x13 -#define DENTIST_DISPCLK_CNTL__DENTIST_DPPCLK_CHG_DONE__SHIFT 0x14 -#define DENTIST_DISPCLK_CNTL__DENTIST_DPPCLK_WDIVIDER__SHIFT 0x18 -#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_WDIVIDER_MASK 0x0000007FL -#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_RDIVIDER_MASK 0x00007F00L -#define DENTIST_DISPCLK_CNTL__DENTIST_DISPCLK_CHG_DONE_MASK 0x00080000L -#define DENTIST_DISPCLK_CNTL__DENTIST_DPPCLK_CHG_DONE_MASK 0x00100000L -#define DENTIST_DISPCLK_CNTL__DENTIST_DPPCLK_WDIVIDER_MASK 0x7F000000L -#define mmDENTIST_DISPCLK_CNTL 0x0124 -#define mmCLK8_CLK_TICK_CNT_CONFIG_REG 0x1B851 -#define mmCLK8_CLK0_CURRENT_CNT 0x1B853 -#define mmCLK8_CLK1_CURRENT_CNT 0x1B854 -#define mmCLK8_CLK2_CURRENT_CNT 0x1B855 -#define mmCLK8_CLK3_CURRENT_CNT 0x1B856 -#define mmCLK8_CLK4_CURRENT_CNT 0x1B857 - - -#define mmCLK8_CLK0_BYPASS_CNTL 0x1B81A -#define mmCLK8_CLK1_BYPASS_CNTL 0x1B822 -#define mmCLK8_CLK2_BYPASS_CNTL 0x1B82A -#define mmCLK8_CLK3_BYPASS_CNTL 0x1B832 -#define mmCLK8_CLK4_BYPASS_CNTL 0x1B83A - - -#define mmCLK8_CLK0_DS_CNTL 0x1B814 -#define mmCLK8_CLK1_DS_CNTL 0x1B81C -#define mmCLK8_CLK2_DS_CNTL 0x1B824 -#define mmCLK8_CLK3_DS_CNTL 0x1B82C -#define mmCLK8_CLK4_DS_CNTL 0x1B834 - - + dc_logger +#define DC_LOGGER_INIT(logger) \ + struct dal_logger *dc_logger = logger +#define DCN42_CLKIP_REFCLK 48000 #undef FN #define FN(reg_name, field_name) \ @@ -92,16 +52,25 @@ #define REG(reg) \ (clk_mgr->regs->reg) +// for DCN register access +#define DCN_BASE__INST0_SEG0 0x00000012 +#define DCN_BASE__INST0_SEG1 0x000000C0 + #define BASE_INNER(seg) DCN_BASE__INST0_SEG ## seg #define BASE(seg) BASE_INNER(seg) -#define SR(reg_name)\ - .reg_name = BASE(reg ## reg_name ## _BASE_IDX) + \ - reg ## reg_name +#define SR(reg_name) \ + .reg_name = BASE(reg ## reg_name ## _BASE_IDX) + reg ## reg_name + +// for CLKIP register access +#define CLK_BASE__INST0_SEG0 0x00016C00 -#define CLK_SR_DCN42(reg_name)\ - .reg_name = mm ## reg_name +#define CLK_BASE_INNER(seg) \ + CLK_BASE__INST0_SEG ## seg + +#define CLK_SR_DCN42(reg_name) \ + .reg_name = CLK_BASE(reg ## reg_name ## _BASE_IDX) + reg ## reg_name static const struct clk_mgr_registers clk_mgr_regs_dcn42 = { CLK_REG_LIST_DCN42() @@ -115,8 +84,6 @@ static const struct clk_mgr_mask clk_mgr_mask_dcn42 = { CLK_COMMON_MASK_SH_LIST_DCN42(_MASK) }; - - #define TO_CLK_MGR_DCN42(clk_mgr_int)\ container_of(clk_mgr_int, struct clk_mgr_dcn42, base) @@ -154,6 +121,49 @@ int dcn42_get_active_display_cnt_wa( return display_count; } +static uint32_t dcn42_get_clock_freq_from_clkip(struct clk_mgr *clk_mgr_base, enum clock_type clock) +{ + struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base); + uint64_t clock_freq_mhz = 0; + uint32_t timer_threshold = 0; + + // always safer to read the timer threshold instead of using cached value + REG_GET(CLK8_CLK_TICK_CNT_CONFIG_REG, TIMER_THRESHOLD, &timer_threshold); + + if (timer_threshold == 0) { + BREAK_TO_DEBUGGER(); + return 0; + } + + switch (clock) { + case clock_type_dispclk: + clock_freq_mhz = REG_READ(CLK8_CLK0_CURRENT_CNT); + break; + case clock_type_dppclk: + clock_freq_mhz = REG_READ(CLK8_CLK1_CURRENT_CNT); + break; + case clock_type_dprefclk: + clock_freq_mhz = REG_READ(CLK8_CLK2_CURRENT_CNT); + break; + case clock_type_dcfclk: + clock_freq_mhz = REG_READ(CLK8_CLK3_CURRENT_CNT); + break; + case clock_type_dtbclk: + clock_freq_mhz = REG_READ(CLK8_CLK4_CURRENT_CNT); + break; + default: + break; + } + + clock_freq_mhz *= DCN42_CLKIP_REFCLK; + clock_freq_mhz = div_u64(clock_freq_mhz, timer_threshold); + + // there are no DCN clocks over 0xFFFFFFFF MHz + ASSERT(clock_freq_mhz <= 0xFFFFFFFF); + + return (uint32_t)clock_freq_mhz; +} + void dcn42_update_clocks_update_dtb_dto(struct clk_mgr_internal *clk_mgr, struct dc_state *context, int ref_dtbclk_khz) @@ -224,8 +234,6 @@ void dcn42_update_clocks(struct clk_mgr *clk_mgr_base, display_count = dcn42_get_active_display_cnt_wa(dc, context, &all_active_disps); - /*dml21 issue*/ - ASSERT(new_clocks->dtbclk_en && new_clocks->ref_dtbclk_khz > 590000); //remove this section if assert is hit if (new_clocks->dtbclk_en && new_clocks->ref_dtbclk_khz < 590000) new_clocks->ref_dtbclk_khz = 600000; @@ -262,9 +270,7 @@ void dcn42_update_clocks(struct clk_mgr *clk_mgr_base, dcn42_update_clocks_update_dtb_dto(clk_mgr, context, new_clocks->ref_dtbclk_khz); dcn42_smu_set_dtbclk(clk_mgr, true); - if (clk_mgr_base->boot_snapshot.timer_threhold) - actual_dtbclk = REG_READ(CLK8_CLK4_CURRENT_CNT) / (clk_mgr_base->boot_snapshot.timer_threhold / 48000); - + actual_dtbclk = dcn42_get_clock_freq_from_clkip(clk_mgr_base, clock_type_dtbclk); if (actual_dtbclk > 590000) { clk_mgr_base->clks.ref_dtbclk_khz = new_clocks->ref_dtbclk_khz; @@ -386,34 +392,27 @@ bool dcn42_are_clock_states_equal(struct dc_clocks *a, static void dcn42_dump_clk_registers_internal(struct dcn42_clk_internal *internal, struct clk_mgr *clk_mgr_base) { struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base); - uint32_t ratio = 1; - - internal->CLK8_CLK_TICK_CNT__TIMER_THRESHOLD = REG_READ(CLK8_CLK_TICK_CNT_CONFIG_REG) & 0xFFFFFF; - - ratio = internal->CLK8_CLK_TICK_CNT__TIMER_THRESHOLD / 48000; - ASSERT(ratio != 0); - - if (ratio) { - // read dcf deep sleep divider - internal->CLK8_CLK0_DS_CNTL = REG_READ(CLK8_CLK0_DS_CNTL); - internal->CLK8_CLK3_DS_CNTL = REG_READ(CLK8_CLK3_DS_CNTL); - // read dispclk - internal->CLK8_CLK0_CURRENT_CNT = REG_READ(CLK8_CLK0_CURRENT_CNT) / ratio; - internal->CLK8_CLK0_BYPASS_CNTL = REG_READ(CLK8_CLK0_BYPASS_CNTL); - // read dppclk - internal->CLK8_CLK1_CURRENT_CNT = REG_READ(CLK8_CLK1_CURRENT_CNT) / ratio; - internal->CLK8_CLK1_BYPASS_CNTL = REG_READ(CLK8_CLK1_BYPASS_CNTL); - // read dprefclk - internal->CLK8_CLK2_CURRENT_CNT = REG_READ(CLK8_CLK2_CURRENT_CNT) / ratio; - internal->CLK8_CLK2_BYPASS_CNTL = REG_READ(CLK8_CLK2_BYPASS_CNTL); - // read dcfclk - internal->CLK8_CLK3_CURRENT_CNT = REG_READ(CLK8_CLK3_CURRENT_CNT) / ratio; - internal->CLK8_CLK3_BYPASS_CNTL = REG_READ(CLK8_CLK3_BYPASS_CNTL); - // read dtbclk - internal->CLK8_CLK4_CURRENT_CNT = REG_READ(CLK8_CLK4_CURRENT_CNT) / ratio; - internal->CLK8_CLK4_BYPASS_CNTL = REG_READ(CLK8_CLK4_BYPASS_CNTL); - } + REG_GET(CLK8_CLK_TICK_CNT_CONFIG_REG, TIMER_THRESHOLD, &internal->CLK8_CLK_TICK_CNT__TIMER_THRESHOLD); + + // read dcf deep sleep divider + internal->CLK8_CLK0_DS_CNTL = REG_READ(CLK8_CLK0_DS_CNTL); + internal->CLK8_CLK3_DS_CNTL = REG_READ(CLK8_CLK3_DS_CNTL); + // read dispclk + internal->CLK8_CLK0_CURRENT_CNT = dcn42_get_clock_freq_from_clkip(clk_mgr_base, clock_type_dispclk); + internal->CLK8_CLK0_BYPASS_CNTL = REG_READ(CLK8_CLK0_BYPASS_CNTL); + // read dppclk + internal->CLK8_CLK1_CURRENT_CNT = dcn42_get_clock_freq_from_clkip(clk_mgr_base, clock_type_dppclk); + internal->CLK8_CLK1_BYPASS_CNTL = REG_READ(CLK8_CLK1_BYPASS_CNTL); + // read dprefclk + internal->CLK8_CLK2_CURRENT_CNT = dcn42_get_clock_freq_from_clkip(clk_mgr_base, clock_type_dprefclk); + internal->CLK8_CLK2_BYPASS_CNTL = REG_READ(CLK8_CLK2_BYPASS_CNTL); + // read dcfclk + internal->CLK8_CLK3_CURRENT_CNT = dcn42_get_clock_freq_from_clkip(clk_mgr_base, clock_type_dcfclk); + internal->CLK8_CLK3_BYPASS_CNTL = REG_READ(CLK8_CLK3_BYPASS_CNTL); + // read dtbclk + internal->CLK8_CLK4_CURRENT_CNT = dcn42_get_clock_freq_from_clkip(clk_mgr_base, clock_type_dtbclk); + internal->CLK8_CLK4_BYPASS_CNTL = REG_READ(CLK8_CLK4_BYPASS_CNTL); } static void dcn42_dump_clk_registers(struct clk_state_registers_and_bypass *regs_and_bypass, @@ -422,8 +421,11 @@ static void dcn42_dump_clk_registers(struct clk_state_registers_and_bypass *regs struct dcn42_clk_internal internal = {0}; char *bypass_clks[5] = {"0x0 DFS", "0x1 REFCLK", "0x2 ERROR", "0x3 400 FCH", "0x4 600 FCH"}; + DC_LOGGER_INIT(clk_mgr->base.base.ctx->logger); + (void)dc_logger; + dcn42_dump_clk_registers_internal(&internal, &clk_mgr->base.base); - regs_and_bypass->timer_threhold = internal.CLK8_CLK_TICK_CNT__TIMER_THRESHOLD; + regs_and_bypass->timer_threshold = internal.CLK8_CLK_TICK_CNT__TIMER_THRESHOLD; regs_and_bypass->dcfclk = internal.CLK8_CLK3_CURRENT_CNT / 10; regs_and_bypass->dcf_deep_sleep_divider = internal.CLK8_CLK3_DS_CNTL / 10; regs_and_bypass->dcf_deep_sleep_allow = internal.CLK8_CLK3_DS_CNTL & 0x10; /*bit 4: CLK0_ALLOW_DS*/ @@ -432,18 +434,10 @@ static void dcn42_dump_clk_registers(struct clk_state_registers_and_bypass *regs regs_and_bypass->dppclk = internal.CLK8_CLK1_CURRENT_CNT / 10; regs_and_bypass->dtbclk = internal.CLK8_CLK4_CURRENT_CNT / 10; - regs_and_bypass->dppclk_bypass = internal.CLK8_CLK1_BYPASS_CNTL & 0x0007; - if (regs_and_bypass->dppclk_bypass > 4) - regs_and_bypass->dppclk_bypass = 0; - regs_and_bypass->dcfclk_bypass = internal.CLK8_CLK3_BYPASS_CNTL & 0x0007; - if (regs_and_bypass->dcfclk_bypass > 4) - regs_and_bypass->dcfclk_bypass = 0; - regs_and_bypass->dispclk_bypass = internal.CLK8_CLK0_BYPASS_CNTL & 0x0007; - if (regs_and_bypass->dispclk_bypass > 4) - regs_and_bypass->dispclk_bypass = 0; - regs_and_bypass->dprefclk_bypass = internal.CLK8_CLK2_BYPASS_CNTL & 0x0007; - if (regs_and_bypass->dprefclk_bypass > 4) - regs_and_bypass->dprefclk_bypass = 0; + regs_and_bypass->dispclk_bypass = get_reg_field_value(internal.CLK8_CLK0_BYPASS_CNTL, CLK8_CLK0_BYPASS_CNTL, CLK0_BYPASS_SEL); + regs_and_bypass->dppclk_bypass = get_reg_field_value(internal.CLK8_CLK1_BYPASS_CNTL, CLK8_CLK1_BYPASS_CNTL, CLK1_BYPASS_SEL); + regs_and_bypass->dprefclk_bypass = get_reg_field_value(internal.CLK8_CLK2_BYPASS_CNTL, CLK8_CLK2_BYPASS_CNTL, CLK2_BYPASS_SEL); + regs_and_bypass->dcfclk_bypass = get_reg_field_value(internal.CLK8_CLK3_BYPASS_CNTL, CLK8_CLK3_BYPASS_CNTL, CLK3_BYPASS_SEL); if (clk_mgr->base.base.ctx->dc->debug.pstate_enabled) { DC_LOG_SMU("clk_type,clk_value,deepsleep_cntl,deepsleep_allow,bypass\n"); @@ -588,6 +582,9 @@ void dcn42_init_clocks(struct clk_mgr *clk_mgr_base) struct clk_mgr_dcn42 *clk_mgr = TO_CLK_MGR_DCN42(clk_mgr_int); struct dcn42_smu_dpm_clks smu_dpm_clks = { 0 }; + DC_LOGGER_INIT(clk_mgr_base->ctx->logger); + (void)dc_logger; + init_clk_states(clk_mgr_base); // to adjust dp_dto reference clock if ssc is enable otherwise to apply dprefclk @@ -1111,25 +1108,30 @@ void dcn42_clk_mgr_construct( */ clk_mgr->base.base.dprefclk_khz = 600000; - clk_mgr->base.smu_present = false; + clk_mgr->base.smu_present = false; + clk_mgr->base.smu_ver = dcn42_smu_get_pmfw_version(&clk_mgr->base); + if (clk_mgr->base.smu_ver && clk_mgr->base.smu_ver != -1) + clk_mgr->base.smu_present = true; - if (ctx->dc_bios->integrated_info) { - clk_mgr->base.base.dentist_vco_freq_khz = ctx->dc_bios->integrated_info->dentist_vco_freq; + if (ctx->dc_bios->integrated_info) { + clk_mgr->base.base.dentist_vco_freq_khz = ctx->dc_bios->integrated_info->dentist_vco_freq; - if (ctx->dc_bios->integrated_info->memory_type == LpDdr5MemType) - dcn42_bw_params.wm_table = lpddr5_wm_table; - else - dcn42_bw_params.wm_table = ddr5_wm_table; - dcn42_bw_params.vram_type = ctx->dc_bios->integrated_info->memory_type; - dcn42_bw_params.dram_channel_width_bytes = ctx->dc_bios->integrated_info->memory_type == 0x22 ? 8 : 4; - dcn42_bw_params.num_channels = ctx->dc_bios->integrated_info->ma_channel_number ? ctx->dc_bios->integrated_info->ma_channel_number : 4; - } - /* in case we don't get a value from the BIOS, use default */ - if (clk_mgr->base.base.dentist_vco_freq_khz == 0) - clk_mgr->base.base.dentist_vco_freq_khz = 3000000; /* 3000MHz */ + if (ctx->dc_bios->integrated_info->memory_type == LpDdr5MemType) + dcn42_bw_params.wm_table = lpddr5_wm_table; + else + dcn42_bw_params.wm_table = ddr5_wm_table; + + dcn42_bw_params.vram_type = ctx->dc_bios->integrated_info->memory_type; + dcn42_bw_params.dram_channel_width_bytes = ctx->dc_bios->integrated_info->memory_type == 0x22 ? 8 : 4; + dcn42_bw_params.num_channels = ctx->dc_bios->integrated_info->ma_channel_number ? ctx->dc_bios->integrated_info->ma_channel_number : 4; + } + + /* in case we don't get a value from the BIOS, use default */ + if (clk_mgr->base.base.dentist_vco_freq_khz == 0) + clk_mgr->base.base.dentist_vco_freq_khz = 3000000; /* 3000MHz */ - /* Saved clocks configured at boot for debug purposes */ - dcn42_dump_clk_registers(&clk_mgr->base.base.boot_snapshot, clk_mgr); + /* Saved clocks configured at boot for debug purposes */ + dcn42_dump_clk_registers(&clk_mgr->base.base.boot_snapshot, clk_mgr); if (clk_mgr->base.smu_present) clk_mgr->base.base.dprefclk_khz = dcn42_smu_get_dprefclk(&clk_mgr->base); diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index 70f1c45ad834..6fbdfb34379d 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -5945,6 +5945,7 @@ bool dc_is_dmub_outbox_supported(struct dc *dc) case AMDGPU_FAMILY_GC_11_0_1: case AMDGPU_FAMILY_GC_11_5_0: + case AMDGPU_FAMILY_GC_11_5_4: if (!dc->debug.dpia_debug.bits.disable_dpia) return true; break; diff --git a/drivers/gpu/drm/amd/display/dc/dc_spl_translate.c b/drivers/gpu/drm/amd/display/dc/dc_spl_translate.c index 37d1a79e8241..ba7bf23f2b2f 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_spl_translate.c +++ b/drivers/gpu/drm/amd/display/dc/dc_spl_translate.c @@ -89,6 +89,7 @@ void translate_SPL_in_params_from_pipe_ctx(struct pipe_ctx *pipe_ctx, struct spl spl_in->callbacks = dcn32_spl_callbacks; break; case DCN_VERSION_4_01: + case DCN_VERSION_4_2: spl_in->callbacks = dcn401_spl_callbacks; break; default: diff --git a/drivers/gpu/drm/amd/display/dc/dccg/dcn20/dcn20_dccg.h b/drivers/gpu/drm/amd/display/dc/dccg/dcn20/dcn20_dccg.h index 463f5826a1ea..088cf305a772 100644 --- a/drivers/gpu/drm/amd/display/dc/dccg/dcn20/dcn20_dccg.h +++ b/drivers/gpu/drm/amd/display/dc/dccg/dcn20/dcn20_dccg.h @@ -374,7 +374,8 @@ type OTG1_DROP_PIXEL;\ type OTG2_DROP_PIXEL;\ type OTG3_ADD_PIXEL;\ - type OTG3_DROP_PIXEL; + type OTG3_DROP_PIXEL;\ + type RESYNC_FIFO_LEVEL_ADJUST_EN; struct dccg_shift { DCCG_REG_FIELD_LIST(uint8_t) diff --git a/drivers/gpu/drm/amd/display/dc/dccg/dcn42/dcn42_dccg.c b/drivers/gpu/drm/amd/display/dc/dccg/dcn42/dcn42_dccg.c index d1593dc68e36..19dfc3fe5c3a 100644 --- a/drivers/gpu/drm/amd/display/dc/dccg/dcn42/dcn42_dccg.c +++ b/drivers/gpu/drm/amd/display/dc/dccg/dcn42/dcn42_dccg.c @@ -180,6 +180,61 @@ void dccg42_set_physymclk( } } +void dccg42_set_pixel_rate_div( + struct dccg *dccg, + uint32_t otg_inst, + enum pixel_rate_div tmds_div, + enum pixel_rate_div unused) +{ + struct dcn_dccg *dccg_dcn = TO_DCN_DCCG(dccg); + uint32_t cur_tmds_div = PIXEL_RATE_DIV_NA; + uint32_t dp_dto_int; + uint32_t reg_val; + + // only 2 and 4 are valid on dcn401 + if (tmds_div != PIXEL_RATE_DIV_BY_2 && tmds_div != PIXEL_RATE_DIV_BY_4) { + return; + } + + dccg401_get_pixel_rate_div(dccg, otg_inst, &cur_tmds_div, &dp_dto_int); + if (tmds_div == cur_tmds_div) + return; + + // encode enum to register value + reg_val = tmds_div == PIXEL_RATE_DIV_BY_4 ? 1 : 0; + + switch (otg_inst) { + case 0: + REG_UPDATE(OTG_PIXEL_RATE_DIV, + OTG0_TMDS_PIXEL_RATE_DIV, reg_val); + break; + case 1: + REG_UPDATE(OTG_PIXEL_RATE_DIV, + OTG1_TMDS_PIXEL_RATE_DIV, reg_val); + break; + case 2: + REG_UPDATE(OTG_PIXEL_RATE_DIV, + OTG2_TMDS_PIXEL_RATE_DIV, reg_val); + break; + case 3: + REG_UPDATE(OTG_PIXEL_RATE_DIV, + OTG3_TMDS_PIXEL_RATE_DIV, reg_val); + break; + default: + BREAK_TO_DEBUGGER(); + return; + } +} + +void dccg42_trigger_dio_fifo_resync(struct dccg *dccg) +{ + struct dcn_dccg *dccg_dcn = TO_DCN_DCCG(dccg); + + REG_UPDATE(DISPCLK_FREQ_CHANGE_CNTL, RESYNC_FIFO_LEVEL_ADJUST_EN, 1); + REG_UPDATE(DISPCLK_FREQ_CHANGE_CNTL, RESYNC_FIFO_LEVEL_ADJUST_EN, 0); + REG_WAIT(DISPCLK_FREQ_CHANGE_CNTL, DISPCLK_FREQ_RAMP_DONE, 1, 50, 2000); +} + static void dccg42_init(struct dccg *dccg) { int otg_inst; @@ -240,9 +295,9 @@ static const struct dccg_funcs dccg42_funcs = { .otg_drop_pixel = dccg42_otg_drop_pixel, .disable_dsc = dccg35_disable_dscclk, .enable_dsc = dccg35_enable_dscclk, - .set_pixel_rate_div = dccg401_set_pixel_rate_div, + .set_pixel_rate_div = dccg42_set_pixel_rate_div, .get_pixel_rate_div = dccg401_get_pixel_rate_div, - .trigger_dio_fifo_resync = dccg35_trigger_dio_fifo_resync, + .trigger_dio_fifo_resync = dccg42_trigger_dio_fifo_resync, .set_dp_dto = dccg401_set_dp_dto, .enable_symclk_se = dccg35_enable_symclk_se, .disable_symclk_se = dccg35_disable_symclk_se, diff --git a/drivers/gpu/drm/amd/display/dc/dccg/dcn42/dcn42_dccg.h b/drivers/gpu/drm/amd/display/dc/dccg/dcn42/dcn42_dccg.h index 96eae0003f43..d9831b0f8235 100644 --- a/drivers/gpu/drm/amd/display/dc/dccg/dcn42/dcn42_dccg.h +++ b/drivers/gpu/drm/amd/display/dc/dccg/dcn42/dcn42_dccg.h @@ -238,7 +238,8 @@ DCCG_SF(SYMCLKE_CLOCK_ENABLE, SYMCLKE_SRC_SEL, mask_sh),\ DCCG_SF(SYMCLKE_CLOCK_ENABLE, SYMCLKE_CLOCK_ENABLE, mask_sh),\ DCCG_SF(SYMCLKE_CLOCK_ENABLE, SYMCLKE_FE_EN, mask_sh),\ - DCCG_SF(SYMCLKE_CLOCK_ENABLE, SYMCLKE_FE_SRC_SEL, mask_sh) + DCCG_SF(SYMCLKE_CLOCK_ENABLE, SYMCLKE_FE_SRC_SEL, mask_sh),\ + DCCG_SF(DISPCLK_FREQ_CHANGE_CNTL, RESYNC_FIFO_LEVEL_ADJUST_EN, mask_sh) void dccg42_otg_add_pixel(struct dccg *dccg, @@ -254,6 +255,14 @@ void dccg42_set_physymclk( enum physymclk_clock_source clk_src, bool force_enable); +void dccg42_set_pixel_rate_div( + struct dccg *dccg, + uint32_t otg_inst, + enum pixel_rate_div tmds_div, + enum pixel_rate_div unused); + +void dccg42_trigger_dio_fifo_resync(struct dccg *dccg); + struct dccg *dccg42_create( struct dc_context *ctx, const struct dccg_registers *regs, diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/bounding_boxes/dcn42_soc_bb.h b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/bounding_boxes/dcn42_soc_bb.h new file mode 100644 index 000000000000..8cd074cfdf18 --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/bounding_boxes/dcn42_soc_bb.h @@ -0,0 +1,359 @@ +/* SPDX-License-Identifier: MIT */ +/* Copyright 2026 Advanced Micro Devices, Inc. */ + +#ifndef __DML_DML_DCN42_SOC_BB__ +#define __DML_DML_DCN42_SOC_BB__ + +#include "dml_top_soc_parameter_types.h" + +static const struct dml2_soc_qos_parameters dml_dcn42_variant_a_soc_qos_params = { + .derate_table = { + .system_active_urgent = { + .dram_derate_percent_pixel = 22, + .dram_derate_percent_vm = 0, + .dram_derate_percent_pixel_and_vm = 0, + .fclk_derate_percent = 76, + .dcfclk_derate_percent = 100, + }, + .system_active_average = { + .dram_derate_percent_pixel = 17, + .dram_derate_percent_vm = 0, + .dram_derate_percent_pixel_and_vm = 0, + .fclk_derate_percent = 57, + .dcfclk_derate_percent = 75, + }, + .dcn_mall_prefetch_urgent = { + .dram_derate_percent_pixel = 40, + .dram_derate_percent_vm = 0, + .dram_derate_percent_pixel_and_vm = 0, + .fclk_derate_percent = 83, + .dcfclk_derate_percent = 100, + }, + .dcn_mall_prefetch_average = { + .dram_derate_percent_pixel = 33, + .dram_derate_percent_vm = 0, + .dram_derate_percent_pixel_and_vm = 0, + .fclk_derate_percent = 62, + .dcfclk_derate_percent = 83, + }, + .system_idle_average = { + .dram_derate_percent_pixel = 70, + .dram_derate_percent_vm = 0, + .dram_derate_percent_pixel_and_vm = 0, + .fclk_derate_percent = 83, + .dcfclk_derate_percent = 100, + }, + }, + .writeback = { + .base_latency_us = 12, + .scaling_factor_us = 0, + .scaling_factor_mhz = 0, + }, + .qos_params = { + .dcn4x = { + .df_qos_response_time_fclk_cycles = 300, + .max_round_trip_to_furthest_cs_fclk_cycles = 350, + .mall_overhead_fclk_cycles = 50, + .meta_trip_adder_fclk_cycles = 36, + .average_transport_distance_fclk_cycles = 257, + .umc_urgent_ramp_latency_margin = 50, + .umc_max_latency_margin = 30, + .umc_average_latency_margin = 20, + .fabric_max_transport_latency_margin = 20, + .fabric_average_transport_latency_margin = 10, + + .per_uclk_dpm_params = { + { + .minimum_uclk_khz = 97 * 1000, + .urgent_ramp_uclk_cycles = 472, + .trip_to_memory_uclk_cycles = 827, + .meta_trip_to_memory_uclk_cycles = 827, + .maximum_latency_when_urgent_uclk_cycles = 72, + .average_latency_when_urgent_uclk_cycles = 61, + .maximum_latency_when_non_urgent_uclk_cycles = 827, + .average_latency_when_non_urgent_uclk_cycles = 118, + }, + }, + }, + }, + .qos_type = dml2_qos_param_type_dcn4x, +}; + +static const struct dml2_soc_bb dml2_socbb_dcn42 = { + .clk_table = { + .uclk = { + .clk_values_khz = {97000}, + .num_clk_values = 1, + }, + .fclk = { + .clk_values_khz = {300000, 2500000}, + .num_clk_values = 2, + }, + .dcfclk = { + .clk_values_khz = {200000, 1564000}, + .num_clk_values = 2, + }, + .dispclk = { + .clk_values_khz = {100000, 1500000}, + .num_clk_values = 2, + }, + .dppclk = { + .clk_values_khz = {100000, 1500000}, + .num_clk_values = 2, + }, + .dtbclk = { + .clk_values_khz = {600000, 600000}, + .num_clk_values = 2, + }, + .phyclk = { + .clk_values_khz = {810000, 810000}, + .num_clk_values = 2, + }, + .socclk = { + .clk_values_khz = {300000, 1200000}, + .num_clk_values = 2, + }, + .dscclk = { + .clk_values_khz = {500000, 500000}, + .num_clk_values = 2, + }, + .phyclk_d18 = { + .clk_values_khz = {667000, 667000}, + .num_clk_values = 2, + }, + .phyclk_d32 = { + .clk_values_khz = {625000, 625000}, + .num_clk_values = 2, + }, + .dram_config = { + .channel_width_bytes = 4, + .channel_count = 4, + .alt_clock_bw_conversion = true, + }, + }, + + .qos_parameters = { + .derate_table = { + .system_active_urgent = { + .dram_derate_percent_pixel = 22, + .dram_derate_percent_vm = 0, + .dram_derate_percent_pixel_and_vm = 0, + .fclk_derate_percent = 76, + .dcfclk_derate_percent = 100, + }, + .system_active_average = { + .dram_derate_percent_pixel = 15, + .dram_derate_percent_vm = 0, + .dram_derate_percent_pixel_and_vm = 0, + .fclk_derate_percent = 57, + .dcfclk_derate_percent = 75, + }, + .dcn_mall_prefetch_urgent = { + .dram_derate_percent_pixel = 40, + .dram_derate_percent_vm = 0, + .dram_derate_percent_pixel_and_vm = 0, + .fclk_derate_percent = 83, + .dcfclk_derate_percent = 100, + }, + .dcn_mall_prefetch_average = { + .dram_derate_percent_pixel = 30, + .dram_derate_percent_vm = 0, + .dram_derate_percent_pixel_and_vm = 0, + .fclk_derate_percent = 62, + .dcfclk_derate_percent = 83, + }, + .system_idle_average = { + .dram_derate_percent_pixel = 70, + .dram_derate_percent_vm = 0, + .dram_derate_percent_pixel_and_vm = 0, + .fclk_derate_percent = 83, + .dcfclk_derate_percent = 100, + }, + }, + .writeback = { + .base_latency_us = 12, + .scaling_factor_us = 0, + .scaling_factor_mhz = 0, + }, + .qos_params = { + .dcn4x = { + .df_qos_response_time_fclk_cycles = 300, + .max_round_trip_to_furthest_cs_fclk_cycles = 350, + .mall_overhead_fclk_cycles = 50, + .meta_trip_adder_fclk_cycles = 36, + .average_transport_distance_fclk_cycles = 260, + .umc_urgent_ramp_latency_margin = 50, + .umc_max_latency_margin = 30, + .umc_average_latency_margin = 20, + .fabric_max_transport_latency_margin = 20, + .fabric_average_transport_latency_margin = 10, + + .per_uclk_dpm_params = { + { + // State 1 + .minimum_uclk_khz = 0, + .urgent_ramp_uclk_cycles = 472, + .trip_to_memory_uclk_cycles = 827, + .meta_trip_to_memory_uclk_cycles = 827, + .maximum_latency_when_urgent_uclk_cycles = 72, + .average_latency_when_urgent_uclk_cycles = 72, + .maximum_latency_when_non_urgent_uclk_cycles = 827, + .average_latency_when_non_urgent_uclk_cycles = 117, + }, + { + // State 2 + .minimum_uclk_khz = 0, + .urgent_ramp_uclk_cycles = 546, + .trip_to_memory_uclk_cycles = 848, + .meta_trip_to_memory_uclk_cycles = 848, + .maximum_latency_when_urgent_uclk_cycles = 146, + .average_latency_when_urgent_uclk_cycles = 146, + .maximum_latency_when_non_urgent_uclk_cycles = 848, + .average_latency_when_non_urgent_uclk_cycles = 133, + }, + { + // State 3 + .minimum_uclk_khz = 0, + .urgent_ramp_uclk_cycles = 564, + .trip_to_memory_uclk_cycles = 853, + .meta_trip_to_memory_uclk_cycles = 853, + .maximum_latency_when_urgent_uclk_cycles = 164, + .average_latency_when_urgent_uclk_cycles = 164, + .maximum_latency_when_non_urgent_uclk_cycles = 853, + .average_latency_when_non_urgent_uclk_cycles = 136, + }, + { + // State 4 + .minimum_uclk_khz = 0, + .urgent_ramp_uclk_cycles = 613, + .trip_to_memory_uclk_cycles = 869, + .meta_trip_to_memory_uclk_cycles = 869, + .maximum_latency_when_urgent_uclk_cycles = 213, + .average_latency_when_urgent_uclk_cycles = 213, + .maximum_latency_when_non_urgent_uclk_cycles = 869, + .average_latency_when_non_urgent_uclk_cycles = 149, + }, + { + // State 5 + .minimum_uclk_khz = 0, + .urgent_ramp_uclk_cycles = 632, + .trip_to_memory_uclk_cycles = 874, + .meta_trip_to_memory_uclk_cycles = 874, + .maximum_latency_when_urgent_uclk_cycles = 232, + .average_latency_when_urgent_uclk_cycles = 232, + .maximum_latency_when_non_urgent_uclk_cycles = 874, + .average_latency_when_non_urgent_uclk_cycles = 153, + }, + { + // State 6 + .minimum_uclk_khz = 0, + .urgent_ramp_uclk_cycles = 665, + .trip_to_memory_uclk_cycles = 885, + .meta_trip_to_memory_uclk_cycles = 885, + .maximum_latency_when_urgent_uclk_cycles = 265, + .average_latency_when_urgent_uclk_cycles = 265, + .maximum_latency_when_non_urgent_uclk_cycles = 885, + .average_latency_when_non_urgent_uclk_cycles = 161, + }, + { + // State 7 + .minimum_uclk_khz = 0, + .urgent_ramp_uclk_cycles = 689, + .trip_to_memory_uclk_cycles = 895, + .meta_trip_to_memory_uclk_cycles = 895, + .maximum_latency_when_urgent_uclk_cycles = 289, + .average_latency_when_urgent_uclk_cycles = 289, + .maximum_latency_when_non_urgent_uclk_cycles = 895, + .average_latency_when_non_urgent_uclk_cycles = 167, + }, + { + // State 8 + .minimum_uclk_khz = 0, + .urgent_ramp_uclk_cycles = 716, + .trip_to_memory_uclk_cycles = 902, + .meta_trip_to_memory_uclk_cycles = 902, + .maximum_latency_when_urgent_uclk_cycles = 316, + .average_latency_when_urgent_uclk_cycles = 316, + .maximum_latency_when_non_urgent_uclk_cycles = 902, + .average_latency_when_non_urgent_uclk_cycles = 174, + }, + }, + }, + }, + .qos_type = dml2_qos_param_type_dcn4x, + }, + + .power_management_parameters = { + .dram_clk_change_blackout_us = 11.72, + .fclk_change_blackout_us = 0, + .g7_ppt_blackout_us = 0, + .stutter_enter_plus_exit_latency_us = 11, + .stutter_exit_latency_us = 9, + .z8_stutter_enter_plus_exit_latency_us = 300, + .z8_stutter_exit_latency_us = 200, + }, + + .vmin_limit = { + .dispclk_khz = 600 * 1000, + }, + + .dprefclk_mhz = 600, + .xtalclk_mhz = 24, + .pcie_refclk_mhz = 100, + .dchub_refclk_mhz = 50, + .mall_allocated_for_dcn_mbytes = 64, + .max_outstanding_reqs = 256, + .fabric_datapath_to_dcn_data_return_bytes = 32, + .return_bus_width_bytes = 64, + .hostvm_min_page_size_kbytes = 4, + .gpuvm_min_page_size_kbytes = 256, + .phy_downspread_percent = 0.38, + .dcn_downspread_percent = 0.38, + .dispclk_dppclk_vco_speed_mhz = 3000, + .do_urgent_latency_adjustment = 0, + .mem_word_bytes = 32, + .num_dcc_mcaches = 8, + .mcache_size_bytes = 2048, + .mcache_line_size_bytes = 32, + .max_fclk_for_uclk_dpm_khz = 2200 * 1000, +}; + +static const struct dml2_ip_capabilities dml2_dcn42_max_ip_caps = { + .pipe_count = 4, + .otg_count = 4, + .num_dsc = 4, + .max_num_dp2p0_streams = 4, + .max_num_hdmi_frl_outputs = 1, + .max_num_dp2p0_outputs = 4, + .rob_buffer_size_kbytes = 64, + .config_return_buffer_size_in_kbytes = 1792, + .config_return_buffer_segment_size_in_kbytes = 64, + .meta_fifo_size_in_kentries = 32, + .compressed_buffer_segment_size_in_kbytes = 64, + .cursor_buffer_size = 24, + .max_flip_time_us = 110, + .max_flip_time_lines = 50, + .hostvm_mode = 0, + .subvp_drr_scheduling_margin_us = 100, + .subvp_prefetch_end_to_mall_start_us = 15, + .subvp_fw_processing_delay = 15, + .max_vactive_det_fill_delay_us = 400, + + .fams2 = { + .max_allow_delay_us = 100 * 1000, + .scheduling_delay_us = 550, + .vertical_interrupt_ack_delay_us = 40, + .allow_programming_delay_us = 18, + .min_allow_width_us = 20, + .subvp_df_throttle_delay_us = 100, + .subvp_programming_delay_us = 200, + .subvp_prefetch_to_mall_delay_us = 18, + .drr_programming_delay_us = 35, + + .lock_timeout_us = 5000, + .recovery_timeout_us = 5000, + .flip_programming_delay_us = 300, + }, +}; + +#endif diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/clk_mgr.h b/drivers/gpu/drm/amd/display/dc/inc/hw/clk_mgr.h index 91eba1985bab..21224fd6b36d 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/clk_mgr.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/clk_mgr.h @@ -215,7 +215,7 @@ struct clk_state_registers_and_bypass { uint32_t dcfclk_bypass; uint32_t dprefclk_bypass; uint32_t dispclk_bypass; - uint32_t timer_threhold; + uint32_t timer_threshold; }; struct rv1_clk_internal { diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/clk_mgr_internal.h b/drivers/gpu/drm/amd/display/dc/inc/hw/clk_mgr_internal.h index 9e53eacee3f8..c69ccfcebeb5 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/clk_mgr_internal.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/clk_mgr_internal.h @@ -243,7 +243,23 @@ enum dentist_divider_range { CLK_SR_DCN42(CLK8_CLK3_DS_CNTL), \ CLK_SR_DCN42(CLK8_CLK4_DS_CNTL) -#define CLK_COMMON_MASK_SH_LIST_DCN42(mask_sh) 0 +#define CLK_COMMON_MASK_SH_LIST_DCN42(mask_sh) \ + CLK_SF(CLK8_CLK_TICK_CNT_CONFIG_REG, TIMER_THRESHOLD, mask_sh), \ + CLK_SF(CLK8_CLK0_BYPASS_CNTL, CLK0_BYPASS_SEL, mask_sh), \ + CLK_SF(CLK8_CLK1_BYPASS_CNTL, CLK1_BYPASS_SEL, mask_sh), \ + CLK_SF(CLK8_CLK2_BYPASS_CNTL, CLK2_BYPASS_SEL, mask_sh), \ + CLK_SF(CLK8_CLK3_BYPASS_CNTL, CLK3_BYPASS_SEL, mask_sh), \ + CLK_SF(CLK8_CLK4_BYPASS_CNTL, CLK4_BYPASS_SEL, mask_sh), \ + CLK_SF(CLK8_CLK0_DS_CNTL, CLK0_DS_DIV_ID, mask_sh), \ + CLK_SF(CLK8_CLK1_DS_CNTL, CLK1_DS_DIV_ID, mask_sh), \ + CLK_SF(CLK8_CLK2_DS_CNTL, CLK2_DS_DIV_ID, mask_sh), \ + CLK_SF(CLK8_CLK3_DS_CNTL, CLK3_DS_DIV_ID, mask_sh), \ + CLK_SF(CLK8_CLK4_DS_CNTL, CLK4_DS_DIV_ID, mask_sh), \ + CLK_SF(CLK8_CLK0_DS_CNTL, CLK0_ALLOW_DS, mask_sh), \ + CLK_SF(CLK8_CLK1_DS_CNTL, CLK1_ALLOW_DS, mask_sh), \ + CLK_SF(CLK8_CLK2_DS_CNTL, CLK2_ALLOW_DS, mask_sh), \ + CLK_SF(CLK8_CLK3_DS_CNTL, CLK3_ALLOW_DS, mask_sh), \ + CLK_SF(CLK8_CLK4_DS_CNTL, CLK4_ALLOW_DS, mask_sh), \ @@ -259,6 +275,42 @@ enum dentist_divider_range { type FbMult_int; \ type FbMult_frac; +#define CLK42_REG_LIST(clkip_num, type) \ + type CLK ## clkip_num ## _CLK_TICK_CNT_CONFIG_REG; \ + type CLK ## clkip_num ## _CLK0_CURRENT_CNT; \ + type CLK ## clkip_num ## _CLK1_CURRENT_CNT; \ + type CLK ## clkip_num ## _CLK2_CURRENT_CNT; \ + type CLK ## clkip_num ## _CLK3_CURRENT_CNT; \ + type CLK ## clkip_num ## _CLK4_CURRENT_CNT; \ + type CLK ## clkip_num ## _CLK0_BYPASS_CNTL; \ + type CLK ## clkip_num ## _CLK1_BYPASS_CNTL; \ + type CLK ## clkip_num ## _CLK2_BYPASS_CNTL; \ + type CLK ## clkip_num ## _CLK3_BYPASS_CNTL; \ + type CLK ## clkip_num ## _CLK4_BYPASS_CNTL; \ + type CLK ## clkip_num ## _CLK0_DS_CNTL; \ + type CLK ## clkip_num ## _CLK1_DS_CNTL; \ + type CLK ## clkip_num ## _CLK2_DS_CNTL; \ + type CLK ## clkip_num ## _CLK3_DS_CNTL; \ + type CLK ## clkip_num ## _CLK4_DS_CNTL; + +#define CLK42_REG_FIELD_LIST(type) \ + type TIMER_THRESHOLD; \ + type CLK0_BYPASS_SEL; \ + type CLK1_BYPASS_SEL; \ + type CLK2_BYPASS_SEL; \ + type CLK3_BYPASS_SEL; \ + type CLK4_BYPASS_SEL; \ + type CLK0_DS_DIV_ID; \ + type CLK1_DS_DIV_ID; \ + type CLK2_DS_DIV_ID; \ + type CLK3_DS_DIV_ID; \ + type CLK4_DS_DIV_ID; \ + type CLK0_ALLOW_DS; \ + type CLK1_ALLOW_DS; \ + type CLK2_ALLOW_DS; \ + type CLK3_ALLOW_DS; \ + type CLK4_ALLOW_DS; + /* *************************************************************************************** ****************** Clock Manager Private Structures *********************************** @@ -322,32 +374,19 @@ struct clk_mgr_registers { uint32_t CLK1_CLK5_ALLOW_DS; uint32_t CLK5_spll_field_8; uint32_t CLK6_spll_field_8; - uint32_t CLK8_CLK0_CURRENT_CNT; - uint32_t CLK8_CLK1_CURRENT_CNT; - uint32_t CLK8_CLK2_CURRENT_CNT; - uint32_t CLK8_CLK3_CURRENT_CNT; - uint32_t CLK8_CLK4_CURRENT_CNT; - uint32_t CLK8_CLK0_DS_CNTL; - uint32_t CLK8_CLK1_DS_CNTL; - uint32_t CLK8_CLK2_DS_CNTL; - uint32_t CLK8_CLK3_DS_CNTL; - uint32_t CLK8_CLK4_DS_CNTL; - uint32_t CLK8_CLK0_BYPASS_CNTL; - uint32_t CLK8_CLK1_BYPASS_CNTL; - uint32_t CLK8_CLK2_BYPASS_CNTL; - uint32_t CLK8_CLK3_BYPASS_CNTL; - uint32_t CLK8_CLK4_BYPASS_CNTL; - uint32_t CLK8_CLK_TICK_CNT_CONFIG_REG; + CLK42_REG_LIST(8, uint32_t) }; struct clk_mgr_shift { CLK_REG_FIELD_LIST(uint8_t) CLK20_REG_FIELD_LIST(uint8_t) + CLK42_REG_FIELD_LIST(uint8_t) }; struct clk_mgr_mask { CLK_REG_FIELD_LIST(uint32_t) CLK20_REG_FIELD_LIST(uint32_t) + CLK42_REG_FIELD_LIST(uint32_t) }; enum clock_type { diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_edp_panel_control.c b/drivers/gpu/drm/amd/display/dc/link/protocols/link_edp_panel_control.c index ba311a8f3f65..4a2699a374b7 100644 --- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_edp_panel_control.c +++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_edp_panel_control.c @@ -847,6 +847,7 @@ bool edp_setup_psr(struct dc_link *link, case FAMILY_YELLOW_CARP: case AMDGPU_FAMILY_GC_10_3_6: case AMDGPU_FAMILY_GC_11_0_1: + case AMDGPU_FAMILY_GC_11_5_4: if (dc->debug.disable_z10 || dc->debug.psr_skip_crtc_disable) psr_context->psr_level.bits.SKIP_CRTC_DISABLE = true; break; diff --git a/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/Makefile b/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/Makefile index bc93356a0b5b..d168fb1eacf7 100644 --- a/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/Makefile +++ b/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/Makefile @@ -8,11 +8,14 @@ soc_and_ip_translator_ccflags := $(CC_FLAGS_FPU) soc_and_ip_translator_rcflags := $(CC_FLAGS_NO_FPU) CFLAGS_$(AMDDALPATH)/dc/soc_and_ip_translator/dcn401/dcn401_soc_and_ip_translator.o := $(soc_and_ip_translator_ccflags) +CFLAGS_$(AMDDALPATH)/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.o := $(soc_and_ip_translator_ccflags) CFLAGS_REMOVE_$(AMDDALPATH)/dc/soc_and_ip_translator/dcn401/dcn401_soc_and_ip_translator.o := $(soc_and_ip_translator_rcflags) +CFLAGS_REMOVE_$(AMDDALPATH)/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.o := $(soc_and_ip_translator_rcflags) soc_and_ip_translator := soc_and_ip_translator.o soc_and_ip_translator += dcn401/dcn401_soc_and_ip_translator.o +soc_and_ip_translator += dcn42/dcn42_soc_and_ip_translator.o AMD_DAL_soc_and_ip_translator := $(addprefix $(AMDDALPATH)/dc/soc_and_ip_translator/, $(soc_and_ip_translator)) diff --git a/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.c b/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.c index c9e224d262c9..87ac56739030 100644 --- a/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.c +++ b/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.c @@ -11,13 +11,169 @@ * This component provides an interface to get DCN-specific bounding box values. */ +static void get_default_soc_bb(struct dml2_soc_bb *soc_bb) +{ + memcpy(soc_bb, &dml2_socbb_dcn42, sizeof(struct dml2_soc_bb)); + memcpy(&soc_bb->qos_parameters, &dml_dcn42_variant_a_soc_qos_params, sizeof(struct dml2_soc_qos_parameters)); +} + +/* + * DC clock table is obtained from SMU during runtime. + * SMU stands for System Management Unit. It is a power management processor. + * It owns the initialization of dc's clock table and programming of clock values + * based on dc's requests. + * Our clock values in base soc bb is a dummy placeholder. The real clock values + * are retrieved from SMU firmware to dc clock table at runtime. + * This function overrides our dummy placeholder values with real values in dc + * clock table. + */ +static void dcn42_convert_dc_clock_table_to_soc_bb_clock_table( + struct dml2_soc_state_table *dml_clk_table, + struct dml2_soc_vmin_clock_limits *vmin_limit, + const struct clk_bw_params *dc_bw_params) +{ + int i; + const struct clk_limit_table *dc_clk_table; + + if (dc_bw_params == NULL) + /* skip if bw params could not be obtained from smu */ + return; + + dc_clk_table = &dc_bw_params->clk_table; + + /* fclk/dcfclk - dcn42 pmfw table can have 0 entries for inactive dpm levels + * for use with dml we need to fill in using an active value aiming for >= 2x DCFCLK + */ + if (dc_clk_table->num_entries_per_clk.num_fclk_levels && dc_clk_table->num_entries_per_clk.num_dcfclk_levels) { + dml_clk_table->fclk.num_clk_values = dc_clk_table->num_entries_per_clk.num_dcfclk_levels; + dml_clk_table->dcfclk.num_clk_values = dc_clk_table->num_entries_per_clk.num_dcfclk_levels; + for (i = 0; i < min(DML_MAX_CLK_TABLE_SIZE, MAX_NUM_DPM_LVL); i++) { + if (i < dc_clk_table->num_entries_per_clk.num_dcfclk_levels) { + int j, max_fclk = 0; + + dml_clk_table->dcfclk.clk_values_khz[i] = dc_clk_table->entries[i].dcfclk_mhz * 1000; + for (j = 0; j < MAX_NUM_DPM_LVL; j++) { + if (dc_clk_table->entries[j].fclk_mhz * 1000 > max_fclk) + max_fclk = dc_clk_table->entries[j].fclk_mhz * 1000; + dml_clk_table->fclk.clk_values_khz[i] = max_fclk; + if (max_fclk >= 2 * dml_clk_table->dcfclk.clk_values_khz[i]) + break; + } + } else { + dml_clk_table->dcfclk.clk_values_khz[i] = 0; + dml_clk_table->fclk.clk_values_khz[i] = 0; + } + } + } + + /* uclk */ + if (dc_clk_table->num_entries_per_clk.num_memclk_levels) { + dml_clk_table->uclk.num_clk_values = dc_clk_table->num_entries_per_clk.num_memclk_levels; + for (i = 0; i < min(DML_MAX_CLK_TABLE_SIZE, MAX_NUM_DPM_LVL); i++) { + if (i < dml_clk_table->uclk.num_clk_values) { + dml_clk_table->uclk.clk_values_khz[i] = dc_clk_table->entries[i].memclk_mhz * 1000; + dml_clk_table->wck_ratio.clk_values_khz[i] = dc_clk_table->entries[i].wck_ratio; + } else { + dml_clk_table->uclk.clk_values_khz[i] = 0; + dml_clk_table->wck_ratio.clk_values_khz[i] = 0; + } + } + } + + /* dispclk */ + if (dc_clk_table->num_entries_per_clk.num_dispclk_levels) { + dml_clk_table->dispclk.num_clk_values = dc_clk_table->num_entries_per_clk.num_dispclk_levels; + for (i = 0; i < min(DML_MAX_CLK_TABLE_SIZE, MAX_NUM_DPM_LVL); i++) { + if (i < dml_clk_table->dispclk.num_clk_values) { + dml_clk_table->dispclk.clk_values_khz[i] = dc_clk_table->entries[i].dispclk_mhz * 1000; + } else { + dml_clk_table->dispclk.clk_values_khz[i] = 0; + } + } + vmin_limit->dispclk_khz = min(dc_clk_table->entries[0].dispclk_mhz * 1000, vmin_limit->dispclk_khz); + } + + /* dppclk */ + if (dc_clk_table->num_entries_per_clk.num_dppclk_levels) { + dml_clk_table->dppclk.num_clk_values = dc_clk_table->num_entries_per_clk.num_dppclk_levels; + for (i = 0; i < min(DML_MAX_CLK_TABLE_SIZE, MAX_NUM_DPM_LVL); i++) { + if (i < dml_clk_table->dppclk.num_clk_values) { + dml_clk_table->dppclk.clk_values_khz[i] = dc_clk_table->entries[i].dppclk_mhz * 1000; + } else { + dml_clk_table->dppclk.clk_values_khz[i] = 0; + } + } + } + + /* dtbclk */ + if (dc_clk_table->num_entries_per_clk.num_dtbclk_levels) { + dml_clk_table->dtbclk.num_clk_values = dc_clk_table->num_entries_per_clk.num_dtbclk_levels; + for (i = 0; i < min(DML_MAX_CLK_TABLE_SIZE, MAX_NUM_DPM_LVL); i++) { + if (i < dml_clk_table->dtbclk.num_clk_values) { + dml_clk_table->dtbclk.clk_values_khz[i] = dc_clk_table->entries[i].dtbclk_mhz * 1000; + } else { + dml_clk_table->dtbclk.clk_values_khz[i] = 0; + } + } + } + + /* socclk */ + if (dc_clk_table->num_entries_per_clk.num_socclk_levels) { + dml_clk_table->socclk.num_clk_values = dc_clk_table->num_entries_per_clk.num_socclk_levels; + for (i = 0; i < min(DML_MAX_CLK_TABLE_SIZE, MAX_NUM_DPM_LVL); i++) { + if (i < dml_clk_table->socclk.num_clk_values) { + dml_clk_table->socclk.clk_values_khz[i] = dc_clk_table->entries[i].socclk_mhz * 1000; + } else { + dml_clk_table->socclk.clk_values_khz[i] = 0; + } + } + } + + /* dram config */ + dml_clk_table->dram_config.channel_count = dc_bw_params->num_channels; + dml_clk_table->dram_config.channel_width_bytes = dc_bw_params->dram_channel_width_bytes; +} + +static void dcn42_update_soc_bb_with_values_from_clk_mgr(struct dml2_soc_bb *soc_bb, const struct dc *dc) +{ + soc_bb->dprefclk_mhz = dc->clk_mgr->dprefclk_khz / 1000; + soc_bb->dispclk_dppclk_vco_speed_mhz = dc->clk_mgr->dentist_vco_freq_khz / 1000.0; + soc_bb->mall_allocated_for_dcn_mbytes = dc->caps.mall_size_total / (1024 * 1024); + + if (dc->clk_mgr->funcs->is_smu_present && + dc->clk_mgr->funcs->is_smu_present(dc->clk_mgr)) { + dcn42_convert_dc_clock_table_to_soc_bb_clock_table(&soc_bb->clk_table, &soc_bb->vmin_limit, + dc->clk_mgr->bw_params); + } +} + +static void apply_soc_bb_updates(struct dml2_soc_bb *soc_bb, const struct dc *dc, const struct dml2_configuration_options *config) +{ + /* Individual modification can be overwritten even if it was obtained by a previous function. + * Modifications are acquired in order of priority (lowest to highest). + */ + dc_assert_fp_enabled(); + + dcn42_update_soc_bb_with_values_from_clk_mgr(soc_bb, dc); + dcn401_update_soc_bb_with_values_from_vbios(soc_bb, dc); + dcn401_update_soc_bb_with_values_from_software_policy(soc_bb, dc); +} + +void dcn42_get_soc_bb(struct dml2_soc_bb *soc_bb, const struct dc *dc, const struct dml2_configuration_options *config) +{ + //get default soc_bb with static values + get_default_soc_bb(soc_bb); + //update soc_bb values with more accurate values + apply_soc_bb_updates(soc_bb, dc, config); +} + static void dcn42_get_ip_caps(struct dml2_ip_capabilities *ip_caps) { *ip_caps = dml2_dcn42_max_ip_caps; } static struct soc_and_ip_translator_funcs dcn42_translator_funcs = { - .get_soc_bb = dcn401_get_soc_bb, + .get_soc_bb = dcn42_get_soc_bb, .get_ip_caps = dcn42_get_ip_caps, }; @@ -25,3 +181,4 @@ void dcn42_construct_soc_and_ip_translator(struct soc_and_ip_translator *soc_and { soc_and_ip_translator->translator_funcs = &dcn42_translator_funcs; } + diff --git a/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.h b/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.h index 914dcbb369a7..1dded5426152 100644 --- a/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.h +++ b/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.h @@ -12,5 +12,6 @@ #include "soc_and_ip_translator.h" void dcn42_construct_soc_and_ip_translator(struct soc_and_ip_translator *soc_and_ip_translator); +void dcn42_get_soc_bb(struct dml2_soc_bb *soc_bb, const struct dc *dc, const struct dml2_configuration_options *config); #endif /* _DCN42_SOC_AND_IP_TRANSLATOR_H_ */ diff --git a/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/soc_and_ip_translator.c b/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/soc_and_ip_translator.c index 6617c9d2d5f8..bad0bd79fa94 100644 --- a/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/soc_and_ip_translator.c +++ b/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/soc_and_ip_translator.c @@ -4,6 +4,7 @@ #include "soc_and_ip_translator.h" #include "soc_and_ip_translator/dcn401/dcn401_soc_and_ip_translator.h" +#include "soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.h" static void dc_construct_soc_and_ip_translator(struct soc_and_ip_translator *soc_and_ip_translator, enum dce_version dc_version) @@ -12,6 +13,9 @@ static void dc_construct_soc_and_ip_translator(struct soc_and_ip_translator *soc case DCN_VERSION_4_01: dcn401_construct_soc_and_ip_translator(soc_and_ip_translator); break; + case DCN_VERSION_4_2: + dcn42_construct_soc_and_ip_translator(soc_and_ip_translator); + break; default: break; } diff --git a/drivers/gpu/drm/amd/include/asic_reg/clk/clk_15_0_0_offset.h b/drivers/gpu/drm/amd/include/asic_reg/clk/clk_15_0_0_offset.h new file mode 100644 index 000000000000..aa3ef3a34013 --- /dev/null +++ b/drivers/gpu/drm/amd/include/asic_reg/clk/clk_15_0_0_offset.h @@ -0,0 +1,44 @@ +/* SPDX-License-Identifier: MIT */ +/* Copyright 2026 Advanced Micro Devices, Inc. */ + +#ifndef _clk_15_0_0_OFFSET_HEADER +#define _clk_15_0_0_OFFSET_HEADER + +// addressBlock: clk_clk8_0_SmuClkDec +// base address: 0x6e000 +#define regCLK8_CLK0_DS_CNTL 0x4c14 +#define regCLK8_CLK0_DS_CNTL_BASE_IDX 0 +#define regCLK8_CLK1_DS_CNTL 0x4c1c +#define regCLK8_CLK1_DS_CNTL_BASE_IDX 0 +#define regCLK8_CLK2_DS_CNTL 0x4c24 +#define regCLK8_CLK2_DS_CNTL_BASE_IDX 0 +#define regCLK8_CLK3_DS_CNTL 0x4c2c +#define regCLK8_CLK3_DS_CNTL_BASE_IDX 0 +#define regCLK8_CLK4_DS_CNTL 0x4c34 +#define regCLK8_CLK4_DS_CNTL_BASE_IDX 0 +#define regCLK8_CLK0_BYPASS_CNTL 0x4c1a +#define regCLK8_CLK0_BYPASS_CNTL_BASE_IDX 0 +#define regCLK8_CLK1_BYPASS_CNTL 0x4c22 +#define regCLK8_CLK1_BYPASS_CNTL_BASE_IDX 0 +#define regCLK8_CLK2_BYPASS_CNTL 0x4c2a +#define regCLK8_CLK2_BYPASS_CNTL_BASE_IDX 0 +#define regCLK8_CLK3_BYPASS_CNTL 0x4c32 +#define regCLK8_CLK3_BYPASS_CNTL_BASE_IDX 0 +#define regCLK8_CLK4_BYPASS_CNTL 0x4c3a +#define regCLK8_CLK4_BYPASS_CNTL_BASE_IDX 0 +#define regCLK8_CLK_TICK_CNT_CONFIG_REG 0x4c51 +#define regCLK8_CLK_TICK_CNT_CONFIG_REG_BASE_IDX 0 +#define regCLK8_CLK_TICK_CNT_STATUS 0x4c52 +#define regCLK8_CLK_TICK_CNT_STATUS_BASE_IDX 0 +#define regCLK8_CLK0_CURRENT_CNT 0x4c53 +#define regCLK8_CLK0_CURRENT_CNT_BASE_IDX 0 +#define regCLK8_CLK1_CURRENT_CNT 0x4c54 +#define regCLK8_CLK1_CURRENT_CNT_BASE_IDX 0 +#define regCLK8_CLK2_CURRENT_CNT 0x4c55 +#define regCLK8_CLK2_CURRENT_CNT_BASE_IDX 0 +#define regCLK8_CLK3_CURRENT_CNT 0x4c56 +#define regCLK8_CLK3_CURRENT_CNT_BASE_IDX 0 +#define regCLK8_CLK4_CURRENT_CNT 0x4c57 +#define regCLK8_CLK4_CURRENT_CNT_BASE_IDX 0 + +#endif diff --git a/drivers/gpu/drm/amd/include/asic_reg/clk/clk_15_0_0_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/clk/clk_15_0_0_sh_mask.h new file mode 100644 index 000000000000..c78622d06e1b --- /dev/null +++ b/drivers/gpu/drm/amd/include/asic_reg/clk/clk_15_0_0_sh_mask.h @@ -0,0 +1,52 @@ +/* SPDX-License-Identifier: MIT */ +/* Copyright 2026 Advanced Micro Devices, Inc. */ + +#ifndef _clk_15_0_0_SH_MASK_HEADER +#define _clk_15_0_0_SH_MASK_HEADER + +// addressBlock: clk_clk8_0_SmuClkDec +//CLK8_CLK_TICK_CNT_CONFIG_REG +#define CLK8_CLK_TICK_CNT_CONFIG_REG__TIMER_THRESHOLD__SHIFT 0x0 +#define CLK8_CLK_TICK_CNT_CONFIG_REG__TIMER_THRESHOLD_MASK 0xFFFFL +//CLK8_CLK0_BYPASS_CNTL +#define CLK8_CLK0_BYPASS_CNTL__CLK0_BYPASS_SEL__SHIFT 0x0 +#define CLK8_CLK0_BYPASS_CNTL__CLK0_BYPASS_SEL_MASK 0x00000007L +//CLK8_CLK1_BYPASS_CNTL +#define CLK8_CLK1_BYPASS_CNTL__CLK1_BYPASS_SEL__SHIFT 0x0 +#define CLK8_CLK1_BYPASS_CNTL__CLK1_BYPASS_SEL_MASK 0x00000007L +//CLK8_CLK2_BYPASS_CNTL +#define CLK8_CLK2_BYPASS_CNTL__CLK2_BYPASS_SEL__SHIFT 0x0 +#define CLK8_CLK2_BYPASS_CNTL__CLK2_BYPASS_SEL_MASK 0x00000007L +//CLK8_CLK3_BYPASS_CNTL +#define CLK8_CLK3_BYPASS_CNTL__CLK3_BYPASS_SEL__SHIFT 0x0 +#define CLK8_CLK3_BYPASS_CNTL__CLK3_BYPASS_SEL_MASK 0x00000007L +//CLK8_CLK4_BYPASS_CNTL +#define CLK8_CLK4_BYPASS_CNTL__CLK4_BYPASS_SEL__SHIFT 0x0 +#define CLK8_CLK4_BYPASS_CNTL__CLK4_BYPASS_SEL_MASK 0x00000007L +//CLK8_CLK0_DS_CNTL +#define CLK8_CLK0_DS_CNTL__CLK0_DS_DIV_ID__SHIFT 0x0 +#define CLK8_CLK0_DS_CNTL__CLK0_DS_DIV_ID_MASK 0x0000000FL +#define CLK8_CLK0_DS_CNTL__CLK0_ALLOW_DS__SHIFT 0x4 +#define CLK8_CLK0_DS_CNTL__CLK0_ALLOW_DS_MASK 0x00000010L +//CLK8_CLK1_DS_CNTL +#define CLK8_CLK1_DS_CNTL__CLK1_DS_DIV_ID__SHIFT 0x0 +#define CLK8_CLK1_DS_CNTL__CLK1_DS_DIV_ID_MASK 0x0000000FL +#define CLK8_CLK1_DS_CNTL__CLK1_ALLOW_DS__SHIFT 0x4 +#define CLK8_CLK1_DS_CNTL__CLK1_ALLOW_DS_MASK 0x00000010L +//CLK8_CLK2_DS_CNTL +#define CLK8_CLK2_DS_CNTL__CLK2_DS_DIV_ID__SHIFT 0x0 +#define CLK8_CLK2_DS_CNTL__CLK2_DS_DIV_ID_MASK 0x0000000FL +#define CLK8_CLK2_DS_CNTL__CLK2_ALLOW_DS__SHIFT 0x4 +#define CLK8_CLK2_DS_CNTL__CLK2_ALLOW_DS_MASK 0x00000010L +//CLK8_CLK3_DS_CNTL +#define CLK8_CLK3_DS_CNTL__CLK3_DS_DIV_ID__SHIFT 0x0 +#define CLK8_CLK3_DS_CNTL__CLK3_DS_DIV_ID_MASK 0x0000000FL +#define CLK8_CLK3_DS_CNTL__CLK3_ALLOW_DS__SHIFT 0x4 +#define CLK8_CLK3_DS_CNTL__CLK3_ALLOW_DS_MASK 0x00000010L +//CLK8_CLK4_DS_CNTL +#define CLK8_CLK4_DS_CNTL__CLK4_DS_DIV_ID__SHIFT 0x0 +#define CLK8_CLK4_DS_CNTL__CLK4_DS_DIV_ID_MASK 0x0000000FL +#define CLK8_CLK4_DS_CNTL__CLK4_ALLOW_DS__SHIFT 0x4 +#define CLK8_CLK4_DS_CNTL__CLK4_ALLOW_DS_MASK 0x00000010L + +#endif -- cgit v1.2.3 From 6e5b72ac22ecedf29d300a6e253296f3a661f7fb Mon Sep 17 00:00:00 2001 From: Roman Li Date: Tue, 3 Mar 2026 12:00:55 -0500 Subject: drm/amd/display: Add DML support for dcn42 DML support for DCN 4.2 Acked-by: Aurabindo Pillai Signed-off-by: Roman Li Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dml2_0/Makefile | 3 + .../dc/dml2_0/dml21/dml21_translation_helper.c | 13 +++ .../display/dc/dml2_0/dml21/inc/dml_top_types.h | 1 + .../dc/dml2_0/dml21/src/dml2_core/dml2_core_dcn4.c | 111 +++++++++++++++++++++ .../dc/dml2_0/dml21/src/dml2_core/dml2_core_dcn4.h | 1 + .../dml2_0/dml21/src/dml2_core/dml2_core_factory.c | 8 ++ .../dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.c | 34 +++++++ .../dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.h | 1 + .../dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_factory.c | 5 + .../dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn42.c | 76 ++++++++++++++ .../dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn42.h | 11 ++ .../dml2_0/dml21/src/dml2_mcg/dml2_mcg_factory.c | 5 + .../dml21/src/dml2_top/dml2_top_interfaces.c | 1 + 13 files changed, 270 insertions(+) create mode 100644 drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn42.c create mode 100644 drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn42.h diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/Makefile b/drivers/gpu/drm/amd/display/dc/dml2_0/Makefile index 30cfc0848792..70d9f2cd0b60 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/Makefile +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/Makefile @@ -90,6 +90,7 @@ CFLAGS_$(AMDDALPATH)/dc/dml2_0/dml21/src/dml2_core/dml2_core_factory.o := $(dml2 CFLAGS_$(AMDDALPATH)/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.o := $(dml2_ccflags) CFLAGS_$(AMDDALPATH)/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_factory.o := $(dml2_ccflags) CFLAGS_$(AMDDALPATH)/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn4.o := $(dml2_ccflags) +CFLAGS_$(AMDDALPATH)/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn42.o := $(dml2_ccflags) CFLAGS_$(AMDDALPATH)/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_factory.o := $(dml2_ccflags) CFLAGS_$(AMDDALPATH)/dc/dml2_0/dml21/src/dml2_pmo/dml2_pmo_dcn3.o := $(dml2_ccflags) CFLAGS_$(AMDDALPATH)/dc/dml2_0/dml21/src/dml2_pmo/dml2_pmo_dcn4_fams2.o := $(dml2_ccflags) @@ -107,6 +108,7 @@ CFLAGS_REMOVE_$(AMDDALPATH)/dc/dml2_0/dml21/src/dml2_top/dml2_top_soc15.o := $(d CFLAGS_REMOVE_$(AMDDALPATH)/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.o := $(dml2_rcflags) CFLAGS_REMOVE_$(AMDDALPATH)/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_factory.o := $(dml2_rcflags) CFLAGS_REMOVE_$(AMDDALPATH)/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn4.o := $(dml2_rcflags) +CFLAGS_REMOVE_$(AMDDALPATH)/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn42.o := $(dml2_rcflags) CFLAGS_REMOVE_$(AMDDALPATH)/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_factory.o := $(dml2_rcflags) CFLAGS_REMOVE_$(AMDDALPATH)/dc/dml2_0/dml21/src/dml2_pmo/dml2_pmo_dcn3.o := $(dml2_rcflags) CFLAGS_REMOVE_$(AMDDALPATH)/dc/dml2_0/dml21/src/dml2_pmo/dml2_pmo_dcn4_fams2.o := $(dml2_rcflags) @@ -124,6 +126,7 @@ DML21 += src/dml2_core/dml2_core_dcn4_calcs.o DML21 += src/dml2_dpmm/dml2_dpmm_dcn4.o DML21 += src/dml2_dpmm/dml2_dpmm_factory.o DML21 += src/dml2_mcg/dml2_mcg_dcn4.o +DML21 += src/dml2_mcg/dml2_mcg_dcn42.o DML21 += src/dml2_mcg/dml2_mcg_factory.o DML21 += src/dml2_pmo/dml2_pmo_dcn3.o DML21 += src/dml2_pmo/dml2_pmo_factory.o diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_translation_helper.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_translation_helper.c index cd12b7a6da65..6cc03391fff6 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_translation_helper.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_translation_helper.c @@ -45,6 +45,9 @@ static enum dml2_project_id dml21_dcn_revision_to_dml2_project_id(enum dce_versi case DCN_VERSION_4_01: project_id = dml2_project_dcn4x_stage2_auto_drr_svp; break; + case DCN_VERSION_4_2: + project_id = dml2_project_dcn42; + break; default: project_id = dml2_project_invalid; DC_ERR("unsupported dcn version for DML21!"); @@ -615,6 +618,13 @@ static void populate_dml21_plane_config_from_plane_state(struct dml2_context *dm case DC_CM2_GPU_MEM_SIZE_171717: plane->tdlut.tdlut_width_mode = dml2_tdlut_width_17_cube; break; + case DC_CM2_GPU_MEM_SIZE_333333: + plane->tdlut.tdlut_width_mode = dml2_tdlut_width_33_cube; + break; + // handling when use case and HW support available + case DC_CM2_GPU_MEM_SIZE_454545: + case DC_CM2_GPU_MEM_SIZE_656565: + break; case DC_CM2_GPU_MEM_SIZE_TRANSFORMED: default: //plane->tdlut.tdlut_width_mode = dml2_tdlut_width_flatten; // dml2_tdlut_width_flatten undefined @@ -824,6 +834,9 @@ void dml21_copy_clocks_to_dc_state(struct dml2_context *in_ctx, struct dc_state context->bw_ctx.bw.dcn.clk.subvp_prefetch_fclk_khz = in_ctx->v21.mode_programming.programming->min_clocks.dcn4x.svp_prefetch_no_throttle.fclk_khz; context->bw_ctx.bw.dcn.clk.stutter_efficiency.base_efficiency = in_ctx->v21.mode_programming.programming->stutter.base_percent_efficiency; context->bw_ctx.bw.dcn.clk.stutter_efficiency.low_power_efficiency = in_ctx->v21.mode_programming.programming->stutter.low_power_percent_efficiency; + context->bw_ctx.bw.dcn.clk.stutter_efficiency.z8_stutter_efficiency = in_ctx->v21.mode_programming.programming->informative.power_management.z8.stutter_efficiency; + context->bw_ctx.bw.dcn.clk.stutter_efficiency.z8_stutter_period = in_ctx->v21.mode_programming.programming->informative.power_management.z8.stutter_period; + context->bw_ctx.bw.dcn.clk.zstate_support = in_ctx->v21.mode_programming.programming->z8_stutter.supported_in_blank; /*ignore meets_eco since it is not used*/ } static struct dml2_dchub_watermark_regs *wm_set_index_to_dc_wm_set(union dcn_watermark_set *watermarks, const enum dml2_dchub_watermark_reg_set_index wm_index) diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/dml_top_types.h b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/dml_top_types.h index 943fd3f040c3..c854515b49a3 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/dml_top_types.h +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/dml_top_types.h @@ -19,6 +19,7 @@ enum dml2_project_id { dml2_project_dcn4x_stage1, dml2_project_dcn4x_stage2, dml2_project_dcn4x_stage2_auto_drr_svp, + dml2_project_dcn42, }; enum dml2_pstate_change_support { diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_dcn4.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_dcn4.c index eba948e187c1..608b4a305c65 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_dcn4.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_dcn4.c @@ -78,6 +78,86 @@ struct dml2_core_ip_params core_dcn4_ip_caps_base = { .subvp_swath_height_margin_lines = 16, }; +struct dml2_core_ip_params core_dcn42_ip_caps_base = { + .vblank_nom_default_us = 668, + .remote_iommu_outstanding_translations = 256, + .rob_buffer_size_kbytes = 64, + .config_return_buffer_size_in_kbytes = 1792, + .config_return_buffer_segment_size_in_kbytes = 64, + .compressed_buffer_segment_size_in_kbytes = 64, + .dpte_buffer_size_in_pte_reqs_luma = 68, + .dpte_buffer_size_in_pte_reqs_chroma = 36, + .pixel_chunk_size_kbytes = 8, + .alpha_pixel_chunk_size_kbytes = 4, + .min_pixel_chunk_size_bytes = 1024, + .writeback_chunk_size_kbytes = 8, + .line_buffer_size_bits = 1171920, + .max_line_buffer_lines = 32, + .writeback_interface_buffer_size_kbytes = 90, + + //Number of pipes after DCN Pipe harvesting + .max_num_dpp = 4, + .max_num_otg = 4, + .max_num_opp = 4, + .max_num_wb = 1, + .max_dchub_pscl_bw_pix_per_clk = 4, + .max_pscl_lb_bw_pix_per_clk = 2, + .max_lb_vscl_bw_pix_per_clk = 4, + .max_vscl_hscl_bw_pix_per_clk = 4, + .max_hscl_ratio = 6, + .max_vscl_ratio = 6, + .max_hscl_taps = 8, + .max_vscl_taps = 8, + .dispclk_ramp_margin_percent = 1, + .dppclk_delay_subtotal = 47, + .dppclk_delay_scl = 50, + .dppclk_delay_scl_lb_only = 16, + .dppclk_delay_cnvc_formatter = 28, + .dppclk_delay_cnvc_cursor = 6, + .cursor_buffer_size = 42, + .cursor_chunk_size = 2, + .dispclk_delay_subtotal = 125, + .max_inter_dcn_tile_repeaters = 8, + .writeback_max_hscl_ratio = 1, + .writeback_max_vscl_ratio = 1, + .writeback_min_hscl_ratio = 1, + .writeback_min_vscl_ratio = 1, + .writeback_max_hscl_taps = 1, + .writeback_max_vscl_taps = 1, + .writeback_line_buffer_buffer_size = 0, + .num_dsc = 4, + .maximum_dsc_bits_per_component = 12, + .maximum_pixels_per_line_per_dsc_unit = 5760, + .dsc422_native_support = true, + .dcc_supported = true, + .ptoi_supported = false, + + .cursor_64bpp_support = true, + .dynamic_metadata_vm_enabled = false, + + .max_num_hdmi_frl_outputs = 0, + .max_num_dp2p0_outputs = 2, + .max_num_dp2p0_streams = 4, + .imall_supported = 1, + .max_flip_time_us = 110, + .max_flip_time_lines = 50, + .words_per_channel = 16, + + .subvp_fw_processing_delay_us = 15, + .subvp_pstate_allow_width_us = 20, + .subvp_swath_height_margin_lines = 16, + + .dcn_mrq_present = 1, + .zero_size_buffer_entries = 512, + .compbuf_reserved_space_zs = 64, + .dcc_meta_buffer_size_bytes = 6272, + .meta_chunk_size_kbytes = 2, + .min_meta_chunk_size_bytes = 256, + + .dchub_arb_to_ret_delay = 102, + .hostvm_mode = 1, +}; + static void patch_ip_caps_with_explicit_ip_params(struct dml2_ip_capabilities *ip_caps, const struct dml2_core_ip_params *ip_params) { ip_caps->pipe_count = ip_params->max_num_dpp; @@ -153,6 +233,37 @@ bool core_dcn4_initialize(struct dml2_core_initialize_in_out *in_out) return true; } +bool core_dcn42_initialize(struct dml2_core_initialize_in_out *in_out) +{ + struct dml2_core_instance *core = in_out->instance; + + if (!in_out->minimum_clock_table) + return false; + else + core->minimum_clock_table = in_out->minimum_clock_table; + + if (in_out->explicit_ip_bb && in_out->explicit_ip_bb_size > 0) { + memcpy(&core->clean_me_up.mode_lib.ip, in_out->explicit_ip_bb, in_out->explicit_ip_bb_size); + + // FIXME_STAGE2: + // DV still uses stage1 ip_param_st for each variant, need to patch the ip_caps with ip_param info + // Should move DV to use ip_caps but need move more overrides to ip_caps + patch_ip_caps_with_explicit_ip_params(in_out->ip_caps, in_out->explicit_ip_bb); + core->clean_me_up.mode_lib.ip.subvp_pstate_allow_width_us = core_dcn4_ip_caps_base.subvp_pstate_allow_width_us; + core->clean_me_up.mode_lib.ip.subvp_fw_processing_delay_us = core_dcn4_ip_caps_base.subvp_pstate_allow_width_us; + core->clean_me_up.mode_lib.ip.subvp_swath_height_margin_lines = core_dcn4_ip_caps_base.subvp_swath_height_margin_lines; + } else { + memcpy(&core->clean_me_up.mode_lib.ip, &core_dcn42_ip_caps_base, sizeof(struct dml2_core_ip_params)); + patch_ip_params_with_ip_caps(&core->clean_me_up.mode_lib.ip, in_out->ip_caps); + core->clean_me_up.mode_lib.ip.imall_supported = false; + } + + memcpy(&core->clean_me_up.mode_lib.soc, in_out->soc_bb, sizeof(struct dml2_soc_bb)); + memcpy(&core->clean_me_up.mode_lib.ip_caps, in_out->ip_caps, sizeof(struct dml2_ip_capabilities)); + + return true; +} + static void create_phantom_stream_from_main_stream(struct dml2_stream_parameters *phantom, const struct dml2_stream_parameters *main, const struct dml2_implicit_svp_meta *meta) { diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_dcn4.h b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_dcn4.h index a68bb001a346..5c26d819a673 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_dcn4.h +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_dcn4.h @@ -5,6 +5,7 @@ #ifndef __DML2_CORE_DCN4_H__ #define __DML2_CORE_DCN4_H__ bool core_dcn4_initialize(struct dml2_core_initialize_in_out *in_out); +bool core_dcn42_initialize(struct dml2_core_initialize_in_out *in_out); bool core_dcn4_mode_support(struct dml2_core_mode_support_in_out *in_out); bool core_dcn4_mode_programming(struct dml2_core_mode_programming_in_out *in_out); bool core_dcn4_populate_informative(struct dml2_core_populate_informative_in_out *in_out); diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_factory.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_factory.c index cc4f0663c6d6..67e307fa4310 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_factory.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_factory.c @@ -30,6 +30,14 @@ bool dml2_core_create(enum dml2_project_id project_id, struct dml2_core_instance out->calculate_mcache_allocation = &core_dcn4_calculate_mcache_allocation; result = true; break; + case dml2_project_dcn42: + out->initialize = &core_dcn42_initialize; + out->mode_support = &core_dcn4_mode_support; + out->mode_programming = &core_dcn4_mode_programming; + out->populate_informative = &core_dcn4_populate_informative; + out->calculate_mcache_allocation = &core_dcn4_calculate_mcache_allocation; + result = true; + break; case dml2_project_invalid: default: break; diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.c index 9d7741fd0adb..037272d2c2e6 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.c @@ -802,3 +802,37 @@ bool dpmm_dcn4_map_watermarks(struct dml2_dpmm_map_watermarks_params_in_out *in_ return true; } + +bool dpmm_dcn42_map_watermarks(struct dml2_dpmm_map_watermarks_params_in_out *in_out) +{ + const struct dml2_display_cfg *display_cfg = &in_out->display_cfg->display_config; + const struct dml2_core_internal_display_mode_lib *mode_lib = &in_out->core->clean_me_up.mode_lib; + struct dml2_dchub_global_register_set *dchubbub_regs = &in_out->programming->global_regs; + + double refclk_freq_in_mhz = (display_cfg->overrides.hw.dlg_ref_clk_mhz > 0) ? (double)display_cfg->overrides.hw.dlg_ref_clk_mhz : mode_lib->soc.dchub_refclk_mhz; + + /* set A */ + dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_A].fclk_pstate = (int unsigned)(mode_lib->mp.Watermark.FCLKChangeWatermark * refclk_freq_in_mhz); + dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_A].sr_enter = (int unsigned)(mode_lib->mp.Watermark.StutterEnterPlusExitWatermark * refclk_freq_in_mhz); + dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_A].sr_exit = (int unsigned)(mode_lib->mp.Watermark.StutterExitWatermark * refclk_freq_in_mhz); + dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_A].sr_enter_z8 = (int unsigned)(mode_lib->mp.Watermark.Z8StutterEnterPlusExitWatermark * refclk_freq_in_mhz); + dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_A].sr_exit_z8 = (int unsigned)(mode_lib->mp.Watermark.Z8StutterExitWatermark * refclk_freq_in_mhz); + dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_A].temp_read_or_ppt = (int unsigned)(mode_lib->mp.Watermark.temp_read_or_ppt_watermark_us * refclk_freq_in_mhz); + dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_A].uclk_pstate = (int unsigned)(mode_lib->mp.Watermark.DRAMClockChangeWatermark * refclk_freq_in_mhz); + dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_A].urgent = (int unsigned)(mode_lib->mp.Watermark.UrgentWatermark * refclk_freq_in_mhz); + dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_A].usr = (int unsigned)(mode_lib->mp.Watermark.USRRetrainingWatermark * refclk_freq_in_mhz); + dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_A].refcyc_per_trip_to_mem = (unsigned int)(mode_lib->mp.Watermark.UrgentWatermark * refclk_freq_in_mhz); + dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_A].refcyc_per_meta_trip_to_mem = (unsigned int)(mode_lib->mp.Watermark.UrgentWatermark * refclk_freq_in_mhz); + dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_A].frac_urg_bw_flip = (unsigned int)(mode_lib->mp.FractionOfUrgentBandwidthImmediateFlip * 1000); + dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_A].frac_urg_bw_nom = (unsigned int)(mode_lib->mp.FractionOfUrgentBandwidth * 1000); + dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_A].frac_urg_bw_mall = (unsigned int)(mode_lib->mp.FractionOfUrgentBandwidthMALL * 1000); + + /* set B */ + dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_B] = dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_A]; + dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_C] = dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_A]; + dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_D] = dchubbub_regs->wm_regs[DML2_DCHUB_WATERMARK_SET_A]; + + dchubbub_regs->num_watermark_sets = 4; + + return true; +} diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.h b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.h index e7b58f2efda4..5fbd07e238a8 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.h +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.h @@ -10,5 +10,6 @@ bool dpmm_dcn3_map_mode_to_soc_dpm(struct dml2_dpmm_map_mode_to_soc_dpm_params_in_out *in_out); bool dpmm_dcn4_map_mode_to_soc_dpm(struct dml2_dpmm_map_mode_to_soc_dpm_params_in_out *in_out); bool dpmm_dcn4_map_watermarks(struct dml2_dpmm_map_watermarks_params_in_out *in_out); +bool dpmm_dcn42_map_watermarks(struct dml2_dpmm_map_watermarks_params_in_out *in_out); #endif diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_factory.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_factory.c index dfd01440737d..80d4932bf154 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_factory.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_factory.c @@ -41,6 +41,11 @@ bool dml2_dpmm_create(enum dml2_project_id project_id, struct dml2_dpmm_instance out->map_watermarks = &dpmm_dcn4_map_watermarks; result = true; break; + case dml2_project_dcn42: + out->map_mode_to_soc_dpm = &dpmm_dcn4_map_mode_to_soc_dpm; + out->map_watermarks = &dpmm_dcn42_map_watermarks; + result = true; + break; case dml2_project_invalid: default: break; diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn42.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn42.c new file mode 100644 index 000000000000..1f67cbc2c236 --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn42.c @@ -0,0 +1,76 @@ +// SPDX-License-Identifier: MIT +// +// Copyright 2026 Advanced Micro Devices, Inc. + +#include "dml2_mcg_dcn42.h" +#include "dml_top_soc_parameter_types.h" + +static unsigned long long uclk_to_dram_bw_kbps(unsigned long uclk_khz, const struct dml2_dram_params *dram_config, unsigned long wck_ratio) +{ + unsigned long long bw_kbps = 0; + + bw_kbps = (unsigned long long) uclk_khz * dram_config->channel_count * dram_config->channel_width_bytes * wck_ratio * 2; + return bw_kbps; +} + +static bool build_min_clk_table_coarse_grained(const struct dml2_soc_bb *soc_bb, struct dml2_mcg_min_clock_table *min_table) +{ + int i; + + for (i = 0; i < soc_bb->clk_table.fclk.num_clk_values; i++) { + if (i < soc_bb->clk_table.uclk.num_clk_values) { + min_table->dram_bw_table.entries[i].pre_derate_dram_bw_kbps = + uclk_to_dram_bw_kbps(soc_bb->clk_table.uclk.clk_values_khz[i], &soc_bb->clk_table.dram_config, soc_bb->clk_table.wck_ratio.clk_values_khz[i]); + min_table->dram_bw_table.entries[i].min_uclk_khz = soc_bb->clk_table.uclk.clk_values_khz[i]; + } else { + min_table->dram_bw_table.entries[i].pre_derate_dram_bw_kbps = min_table->dram_bw_table.entries[soc_bb->clk_table.uclk.num_clk_values - 1].pre_derate_dram_bw_kbps; + min_table->dram_bw_table.entries[i].min_uclk_khz = soc_bb->clk_table.uclk.clk_values_khz[soc_bb->clk_table.uclk.num_clk_values - 1]; + } + + min_table->dram_bw_table.entries[i].min_dcfclk_khz = soc_bb->clk_table.dcfclk.clk_values_khz[i]; + min_table->dram_bw_table.entries[i].min_fclk_khz = soc_bb->clk_table.fclk.clk_values_khz[i]; + } + min_table->dram_bw_table.num_entries = soc_bb->clk_table.fclk.num_clk_values; + + return true; +} + +static bool build_min_clock_table(const struct dml2_soc_bb *soc_bb, struct dml2_mcg_min_clock_table *min_table) +{ + bool result; + + if (!soc_bb || !min_table) + return false; + + + if (soc_bb->clk_table.uclk.num_clk_values > DML_MCG_MAX_CLK_TABLE_SIZE) + return false; + + min_table->fixed_clocks_khz.amclk = 0; + min_table->fixed_clocks_khz.dprefclk = soc_bb->dprefclk_mhz * 1000; + min_table->fixed_clocks_khz.pcierefclk = soc_bb->pcie_refclk_mhz * 1000; + min_table->fixed_clocks_khz.dchubrefclk = soc_bb->dchub_refclk_mhz * 1000; + min_table->fixed_clocks_khz.xtalclk = soc_bb->xtalclk_mhz * 1000; + + min_table->max_clocks_khz.dispclk = soc_bb->clk_table.dispclk.clk_values_khz[soc_bb->clk_table.dispclk.num_clk_values - 1]; + min_table->max_clocks_khz.dppclk = soc_bb->clk_table.dppclk.clk_values_khz[soc_bb->clk_table.dppclk.num_clk_values - 1]; + min_table->max_clocks_khz.dscclk = soc_bb->clk_table.dscclk.clk_values_khz[soc_bb->clk_table.dscclk.num_clk_values - 1]; + min_table->max_clocks_khz.dtbclk = soc_bb->clk_table.dtbclk.clk_values_khz[soc_bb->clk_table.dtbclk.num_clk_values - 1]; + min_table->max_clocks_khz.phyclk = soc_bb->clk_table.phyclk.clk_values_khz[soc_bb->clk_table.phyclk.num_clk_values - 1]; + + min_table->max_ss_clocks_khz.dispclk = (unsigned int)((double)min_table->max_clocks_khz.dispclk / (1.0 + soc_bb->dcn_downspread_percent / 100.0)); + min_table->max_ss_clocks_khz.dppclk = (unsigned int)((double)min_table->max_clocks_khz.dppclk / (1.0 + soc_bb->dcn_downspread_percent / 100.0)); + min_table->max_ss_clocks_khz.dtbclk = (unsigned int)((double)min_table->max_clocks_khz.dtbclk / (1.0 + soc_bb->dcn_downspread_percent / 100.0)); + + min_table->max_clocks_khz.dcfclk = soc_bb->clk_table.dcfclk.clk_values_khz[soc_bb->clk_table.dcfclk.num_clk_values - 1]; + min_table->max_clocks_khz.fclk = soc_bb->clk_table.fclk.clk_values_khz[soc_bb->clk_table.fclk.num_clk_values - 1]; + + result = build_min_clk_table_coarse_grained(soc_bb, min_table); + + return result; +} + +bool mcg_dcn42_build_min_clock_table(struct dml2_mcg_build_min_clock_table_params_in_out *in_out) +{ + return build_min_clock_table(in_out->soc_bb, in_out->min_clk_table); +} diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn42.h b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn42.h new file mode 100644 index 000000000000..ad8a847a0a17 --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn42.h @@ -0,0 +1,11 @@ +/* SPDX-License-Identifier: MIT */ +/* Copyright 2026 Advanced Micro Devices, Inc. */ + +#ifndef __DML2_MCG_DCN42_H__ +#define __DML2_MCG_DCN42_H__ + +#include "dml2_internal_shared_types.h" + +bool mcg_dcn42_build_min_clock_table(struct dml2_mcg_build_min_clock_table_params_in_out *in_out); + +#endif diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_factory.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_factory.c index c60b8fe90819..1cff62b63e72 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_factory.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_factory.c @@ -4,6 +4,7 @@ #include "dml2_mcg_factory.h" #include "dml2_mcg_dcn4.h" +#include "dml2_mcg_dcn42.h" #include "dml2_external_lib_deps.h" static bool dummy_build_min_clock_table(struct dml2_mcg_build_min_clock_table_params_in_out *in_out) @@ -30,6 +31,10 @@ bool dml2_mcg_create(enum dml2_project_id project_id, struct dml2_mcg_instance * out->build_min_clock_table = &mcg_dcn4_build_min_clock_table; result = true; break; + case dml2_project_dcn42: + out->build_min_clock_table = &mcg_dcn42_build_min_clock_table; + result = true; + break; case dml2_project_invalid: default: break; diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_top/dml2_top_interfaces.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_top/dml2_top_interfaces.c index 5a33e2f357f4..04860b6790df 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_top/dml2_top_interfaces.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_top/dml2_top_interfaces.c @@ -17,6 +17,7 @@ bool dml2_initialize_instance(struct dml2_initialize_instance_in_out *in_out) case dml2_project_dcn4x_stage1: case dml2_project_dcn4x_stage2: case dml2_project_dcn4x_stage2_auto_drr_svp: + case dml2_project_dcn42: return dml2_top_soc15_initialize_instance(in_out); case dml2_project_invalid: default: -- cgit v1.2.3 From e56e3cff2a1bb29545ddbec562e76c0419363a40 Mon Sep 17 00:00:00 2001 From: Roman Li Date: Thu, 5 Mar 2026 12:56:09 -0500 Subject: drm/amd/display: Sync dcn42 with DC 3.2.373 This patch provides a bulk merge to align driver support for DCN42 with Display Core version 3.2.373. It includes upgrade for: - clk_mgr - dml2/dml21 - optc - hubp - mpc - optc - hwseq Acked-by: Aurabindo Pillai Signed-off-by: Roman Li Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/bios/bios_parser.c | 11 +- .../amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c | 141 +++--- .../amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.h | 2 +- drivers/gpu/drm/amd/display/dc/core/dc.c | 95 ++-- drivers/gpu/drm/amd/display/dc/core/dc_stream.c | 41 +- drivers/gpu/drm/amd/display/dc/core/dc_surface.c | 9 + drivers/gpu/drm/amd/display/dc/dc.h | 59 ++- drivers/gpu/drm/amd/display/dc/dc_bios_types.h | 3 +- drivers/gpu/drm/amd/display/dc/dc_plane.h | 1 + drivers/gpu/drm/amd/display/dc/dc_types.h | 24 + .../drm/amd/display/dc/dccg/dcn401/dcn401_dccg.c | 20 + .../gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.c | 16 + .../gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.h | 13 + .../dc/dml2_0/dml21/dml21_translation_helper.c | 30 +- .../drm/amd/display/dc/dml2_0/dml21/dml21_utils.c | 1 + .../amd/display/dc/dml2_0/dml21/dml21_wrapper.c | 14 +- .../dml2_0/dml21/inc/bounding_boxes/dcn42_soc_bb.h | 306 ++++-------- .../dml2_0/dml21/inc/dml_top_display_cfg_types.h | 13 + .../display/dc/dml2_0/dml21/inc/dml_top_types.h | 2 + .../dml2_0/dml21/src/dml2_core/dml2_core_factory.c | 1 + .../dml2_0/dml21/src/dml2_core/dml2_core_utils.c | 63 ++- .../dml2_0/dml21/src/dml2_core/dml2_core_utils.h | 2 + .../dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.c | 1 - .../dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_factory.c | 1 + .../dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn4.h | 2 +- .../dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn42.h | 3 +- .../dml2_0/dml21/src/dml2_mcg/dml2_mcg_factory.c | 1 + .../dml21/src/dml2_pmo/dml2_pmo_dcn4_fams2.c | 6 + .../dml2_0/dml21/src/dml2_pmo/dml2_pmo_factory.c | 4 +- .../dml2_0/dml21/src/dml2_pmo/dml2_pmo_factory.h | 2 +- .../dml21/src/dml2_top/dml2_top_interfaces.c | 1 + .../dc/dml2_0/dml21/src/dml2_top/dml2_top_legacy.c | 1 - .../dml21/src/inc/dml2_internal_shared_types.h | 3 + .../gpu/drm/amd/display/dc/dml2_0/dml2_wrapper.c | 3 +- .../drm/amd/display/dc/dpp/dcn30/dcn30_dpp_cm.c | 2 + .../amd/display/dc/dpp/dcn401/dcn401_dpp_dscl.c | 23 +- .../dc/hpo/dcn31/dcn31_hpo_dp_stream_encoder.c | 5 + .../drm/amd/display/dc/hubp/dcn401/dcn401_hubp.c | 201 +++++--- .../drm/amd/display/dc/hubp/dcn401/dcn401_hubp.h | 23 +- .../gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.c | 85 +++- .../gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.h | 16 +- .../drm/amd/display/dc/hwss/dce110/dce110_hwseq.c | 136 ++--- .../drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c | 412 ++++++--------- .../drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.c | 553 ++++----------------- .../drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.h | 10 +- .../gpu/drm/amd/display/dc/hwss/dcn42/dcn42_init.c | 8 +- drivers/gpu/drm/amd/display/dc/inc/hw/hubp.h | 36 +- drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h | 18 +- drivers/gpu/drm/amd/display/dc/inc/hw/mpc.h | 112 +++-- drivers/gpu/drm/amd/display/dc/inc/resource.h | 1 + .../gpu/drm/amd/display/dc/mpc/dcn401/dcn401_mpc.c | 177 ++++--- .../gpu/drm/amd/display/dc/mpc/dcn401/dcn401_mpc.h | 25 +- .../gpu/drm/amd/display/dc/mpc/dcn42/dcn42_mpc.c | 390 ++------------- .../gpu/drm/amd/display/dc/mpc/dcn42/dcn42_mpc.h | 50 +- .../gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.h | 2 +- .../gpu/drm/amd/display/dc/optc/dcn42/dcn42_optc.c | 105 +++- .../gpu/drm/amd/display/dc/optc/dcn42/dcn42_optc.h | 13 +- .../amd/display/dc/resource/dcn42/dcn42_resource.c | 10 +- .../dcn401/dcn401_soc_and_ip_translator.c | 3 + .../dcn42/dcn42_soc_and_ip_translator.c | 12 +- drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h | 38 +- .../amd/include/asic_reg/dcn/dcn_4_2_0_offset.h | 2 + .../amd/include/asic_reg/dcn/dcn_4_2_0_sh_mask.h | 9 + 63 files changed, 1493 insertions(+), 1879 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/bios/bios_parser.c b/drivers/gpu/drm/amd/display/dc/bios/bios_parser.c index 242032c047ed..f947f82013c6 100644 --- a/drivers/gpu/drm/amd/display/dc/bios/bios_parser.c +++ b/drivers/gpu/drm/amd/display/dc/bios/bios_parser.c @@ -794,13 +794,11 @@ static enum bp_result bios_parser_external_encoder_control( static enum bp_result bios_parser_dac_load_detection( struct dc_bios *dcb, - enum engine_id engine_id, - struct graphics_object_id ext_enc_id) + enum engine_id engine_id) { struct bios_parser *bp = BP_FROM_DCB(dcb); struct dc_context *ctx = dcb->ctx; struct bp_load_detection_parameters bp_params = {0}; - struct bp_external_encoder_control ext_cntl = {0}; enum bp_result bp_result = BP_RESULT_UNSUPPORTED; uint32_t bios_0_scratch; uint32_t device_id_mask = 0; @@ -826,13 +824,6 @@ static enum bp_result bios_parser_dac_load_detection( bp_params.engine_id = engine_id; bp_result = bp->cmd_tbl.dac_load_detection(bp, &bp_params); - } else if (ext_enc_id.id) { - if (!bp->cmd_tbl.external_encoder_control) - return BP_RESULT_UNSUPPORTED; - - ext_cntl.action = EXTERNAL_ENCODER_CONTROL_DAC_LOAD_DETECT; - ext_cntl.encoder_id = ext_enc_id; - bp_result = bp->cmd_tbl.external_encoder_control(bp, &ext_cntl); } if (bp_result != BP_RESULT_OK) diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c index cbb79789a532..97c9f0ce13e4 100644 --- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c +++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.c @@ -87,21 +87,18 @@ static const struct clk_mgr_mask clk_mgr_mask_dcn42 = { #define TO_CLK_MGR_DCN42(clk_mgr_int)\ container_of(clk_mgr_int, struct clk_mgr_dcn42, base) -int dcn42_get_active_display_cnt_wa( - struct dc *dc, - struct dc_state *context, - int *all_active_disps) +bool dcn42_has_active_display(struct dc *dc, const struct dc_state *context) { - int i, display_count = 0; - bool tmds_present = false; + int i, active_count = 0; for (i = 0; i < context->stream_count; i++) { const struct dc_stream_state *stream = context->streams[i]; - if (stream->signal == SIGNAL_TYPE_HDMI_TYPE_A || - stream->signal == SIGNAL_TYPE_DVI_SINGLE_LINK || - stream->signal == SIGNAL_TYPE_DVI_DUAL_LINK) - tmds_present = true; + /* Checking stream / link detection ensuring that PHY is active*/ + if (dc_is_hdmi_signal(stream->signal) || + dc_is_dvi_signal(stream->signal) || + (dc_is_dp_signal(stream->signal) && !stream->dpms_off)) + active_count++; } for (i = 0; i < dc->link_count; i++) { @@ -110,15 +107,10 @@ int dcn42_get_active_display_cnt_wa( /* abusing the fact that the dig and phy are coupled to see if the phy is enabled */ if (link->link_enc && link->link_enc->funcs->is_dig_enabled && link->link_enc->funcs->is_dig_enabled(link->link_enc)) - display_count++; + active_count++; } - if (all_active_disps != NULL) - *all_active_disps = display_count; - /* WA for hang on HDMI after display off back on*/ - if (display_count == 0 && tmds_present) - display_count = 1; - return display_count; + return active_count > 0; } static uint32_t dcn42_get_clock_freq_from_clkip(struct clk_mgr *clk_mgr_base, enum clock_type clock) @@ -223,20 +215,18 @@ void dcn42_update_clocks(struct clk_mgr *clk_mgr_base, struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base); struct dc_clocks *new_clocks = &context->bw_ctx.bw.dcn.clk; struct dc *dc = clk_mgr_base->ctx->dc; - int display_count = 0; bool update_dppclk = false; bool update_dispclk = false; bool dpp_clock_lowered = false; - int all_active_disps = 0; + bool has_active_display; if (dc->work_arounds.skip_clock_update) return; - display_count = dcn42_get_active_display_cnt_wa(dc, context, &all_active_disps); + has_active_display = dcn42_has_active_display(dc, context); if (new_clocks->dtbclk_en && new_clocks->ref_dtbclk_khz < 590000) new_clocks->ref_dtbclk_khz = 600000; - /* * if it is safe to lower, but we are already in the lower state, we don't have to do anything * also if safe to lower is false, we just go in the higher state @@ -256,7 +246,7 @@ void dcn42_update_clocks(struct clk_mgr *clk_mgr_base, /* check that we're not already in lower */ if (clk_mgr_base->clks.pwr_state != DCN_PWR_STATE_LOW_POWER) { /* if we can go lower, go lower */ - if (display_count == 0) + if (has_active_display == false) clk_mgr_base->clks.pwr_state = DCN_PWR_STATE_LOW_POWER; } } else { @@ -314,7 +304,7 @@ void dcn42_update_clocks(struct clk_mgr *clk_mgr_base, } if (should_set_clock(safe_to_lower, new_clocks->dispclk_khz, clk_mgr_base->clks.dispclk_khz) && - (new_clocks->dispclk_khz > 0 || (safe_to_lower && display_count == 0))) { + (new_clocks->dispclk_khz > 0 || (safe_to_lower && has_active_display == false))) { int requested_dispclk_khz = new_clocks->dispclk_khz; dcn35_disable_otg_wa(clk_mgr_base, context, safe_to_lower, true); @@ -461,7 +451,6 @@ static void dcn42_dump_clk_registers(struct clk_state_registers_and_bypass *regs // REGISTER VALUES DC_LOG_SMU("reg_name,value,clk_type\n"); - DC_LOG_SMU("CLK1_CLK3_CURRENT_CNT,%d,dcfclk\n", internal.CLK8_CLK3_CURRENT_CNT); @@ -594,6 +583,7 @@ void dcn42_init_clocks(struct clk_mgr *clk_mgr_base) else clk_mgr_base->dp_dto_source_clock_in_khz = clk_mgr_base->dprefclk_khz; + DC_LOG_SMU("dp_dto_source_clock %d, dprefclk %d\n", clk_mgr_base->dp_dto_source_clock_in_khz, clk_mgr_base->dprefclk_khz); dcn42_dump_clk_registers(&clk_mgr_base->boot_snapshot, clk_mgr); clk_mgr_base->clks.ref_dtbclk_khz = clk_mgr_base->boot_snapshot.dtbclk * 10; @@ -602,6 +592,12 @@ void dcn42_init_clocks(struct clk_mgr *clk_mgr_base) clk_mgr_base->clks.dtbclk_en = true; } + if (clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_dcfclk_levels != 0) { + /*skip to get clock table and notify pmfw watermark range again*/ + DC_LOG_SMU("skip to get dpm_clks from pmfw from resume and acr\n"); + return; + } + smu_dpm_clks.dpm_clks = (DpmClocks_t_dcn42 *)dm_helpers_allocate_gpu_mem( clk_mgr_base->ctx, DC_MEM_ALLOC_TYPE_GART, @@ -708,10 +704,9 @@ void dcn42_init_clocks(struct clk_mgr *clk_mgr_base) /* DTBCLK*/ clk_mgr_base->bw_params->clk_table.entries[0].dtbclk_mhz = clk_mgr_base->clks.ref_dtbclk_khz / 1000; clk_mgr_base->bw_params->clk_table.num_entries_per_clk.num_dtbclk_levels = 1; - /* Refresh bounding box */ clk_mgr_base->ctx->dc->res_pool->funcs->update_bw_bounding_box( - clk_mgr_base->ctx->dc, clk_mgr_base->bw_params); + clk_mgr_base->ctx->dc, clk_mgr_base->bw_params); } } if (smu_dpm_clks.dpm_clks && smu_dpm_clks.mc_address.quad_part != 0) @@ -823,7 +818,6 @@ static void dcn42_read_ss_info_from_lut(struct clk_mgr_internal *clk_mgr) } } -/* Exposed for dcn42b reuse */ void dcn42_build_watermark_ranges(struct clk_bw_params *bw_params, struct dcn42_watermarks *table) { int i, num_valid_sets; @@ -882,18 +876,42 @@ void dcn42_build_watermark_ranges(struct clk_bw_params *bw_params, struct dcn42_ void dcn42_notify_wm_ranges(struct clk_mgr *clk_mgr_base) { + int i = 0; + struct dcn42_watermarks *table = NULL; struct clk_mgr_internal *clk_mgr = TO_CLK_MGR_INTERNAL(clk_mgr_base); struct clk_mgr_dcn42 *clk_mgr_dcn42 = TO_CLK_MGR_DCN42(clk_mgr); - struct dcn42_watermarks *table = clk_mgr_dcn42->smu_wm_set.wm_set; if (!clk_mgr->smu_ver) return; + /*send once already skip*/ + if (clk_mgr_base->bw_params->wm_table.entries[WM_A].valid == true) + return; + clk_mgr_dcn42->smu_wm_set.wm_set = (struct dcn42_watermarks *)dm_helpers_allocate_gpu_mem( + clk_mgr->base.ctx, + DC_MEM_ALLOC_TYPE_GART, + sizeof(struct dcn42_watermarks), + &clk_mgr_dcn42->smu_wm_set.mc_address.quad_part); + + ASSERT(clk_mgr_dcn42->smu_wm_set.wm_set); + + table = clk_mgr_dcn42->smu_wm_set.wm_set; if (!table || clk_mgr_dcn42->smu_wm_set.mc_address.quad_part == 0) return; memset(table, 0, sizeof(*table)); + /*same as previous asic, set wm valid before building watermark ranges*/ + for (i = 0; i < WM_SET_COUNT; i++) { + clk_mgr_base->bw_params->wm_table.entries[i].wm_inst = i; + if (i >= clk_mgr_base->bw_params->clk_table.num_entries) { + clk_mgr_base->bw_params->wm_table.entries[i].valid = false; + continue; + } + clk_mgr_base->bw_params->wm_table.entries[i].wm_type = WM_TYPE_PSTATE_CHG; + clk_mgr_base->bw_params->wm_table.entries[i].valid = true; + } + /* build watermark_range will check this valid range*/ dcn42_build_watermark_ranges(clk_mgr_base->bw_params, table); dcn42_smu_set_dram_addr_high(clk_mgr, @@ -901,18 +919,21 @@ void dcn42_notify_wm_ranges(struct clk_mgr *clk_mgr_base) dcn42_smu_set_dram_addr_low(clk_mgr, clk_mgr_dcn42->smu_wm_set.mc_address.low_part); dcn42_smu_transfer_wm_table_dram_2_smu(clk_mgr); + + if (clk_mgr_dcn42->smu_wm_set.wm_set && clk_mgr_dcn42->smu_wm_set.mc_address.quad_part != 0) + dm_helpers_free_gpu_mem(clk_mgr->base.ctx, DC_MEM_ALLOC_TYPE_GART, + clk_mgr_dcn42->smu_wm_set.wm_set); + } void dcn42_set_low_power_state(struct clk_mgr *clk_mgr_base) { - int display_count; struct dc *dc = clk_mgr_base->ctx->dc; struct dc_state *context = dc->current_state; if (clk_mgr_base->clks.pwr_state != DCN_PWR_STATE_LOW_POWER) { - display_count = dcn42_get_active_display_cnt_wa(dc, context, NULL); /* if we can go lower, go lower */ - if (display_count == 0) + if (dcn42_has_active_display(dc, context) == false) clk_mgr_base->clks.pwr_state = DCN_PWR_STATE_LOW_POWER; } @@ -1093,14 +1114,7 @@ void dcn42_clk_mgr_construct( clk_mgr->base.dprefclk_ss_divider = 1000; clk_mgr->base.ss_on_dprefclk = false; clk_mgr->base.dfs_ref_freq_khz = 48000; /*sync with pmfw*/ - - clk_mgr->smu_wm_set.wm_set = (struct dcn42_watermarks *)dm_helpers_allocate_gpu_mem( - clk_mgr->base.base.ctx, - DC_MEM_ALLOC_TYPE_GART, - sizeof(struct dcn42_watermarks), - &clk_mgr->smu_wm_set.mc_address.quad_part); - - ASSERT(clk_mgr->smu_wm_set.wm_set); + clk_mgr->base.base.clks.ref_dtbclk_khz = 600000; /* Changed from DCN3.2_clock_frequency doc to match * dcn32_dump_clk_registers from 4 * dentist_vco_freq_khz / @@ -1108,34 +1122,31 @@ void dcn42_clk_mgr_construct( */ clk_mgr->base.base.dprefclk_khz = 600000; - clk_mgr->base.smu_present = false; - clk_mgr->base.smu_ver = dcn42_smu_get_pmfw_version(&clk_mgr->base); - if (clk_mgr->base.smu_ver && clk_mgr->base.smu_ver != -1) - clk_mgr->base.smu_present = true; - - if (ctx->dc_bios->integrated_info) { - clk_mgr->base.base.dentist_vco_freq_khz = ctx->dc_bios->integrated_info->dentist_vco_freq; - - if (ctx->dc_bios->integrated_info->memory_type == LpDdr5MemType) - dcn42_bw_params.wm_table = lpddr5_wm_table; - else - dcn42_bw_params.wm_table = ddr5_wm_table; - - dcn42_bw_params.vram_type = ctx->dc_bios->integrated_info->memory_type; - dcn42_bw_params.dram_channel_width_bytes = ctx->dc_bios->integrated_info->memory_type == 0x22 ? 8 : 4; - dcn42_bw_params.num_channels = ctx->dc_bios->integrated_info->ma_channel_number ? ctx->dc_bios->integrated_info->ma_channel_number : 4; - } - - /* in case we don't get a value from the BIOS, use default */ - if (clk_mgr->base.base.dentist_vco_freq_khz == 0) - clk_mgr->base.base.dentist_vco_freq_khz = 3000000; /* 3000MHz */ + clk_mgr->base.smu_present = false; + clk_mgr->base.smu_ver = dcn42_smu_get_pmfw_version(&clk_mgr->base); + if (clk_mgr->base.smu_ver && clk_mgr->base.smu_ver != -1) + clk_mgr->base.smu_present = true; + + if (ctx->dc_bios->integrated_info) { + clk_mgr->base.base.dentist_vco_freq_khz = ctx->dc_bios->integrated_info->dentist_vco_freq; + + if (ctx->dc_bios->integrated_info->memory_type == LpDdr5MemType) + dcn42_bw_params.wm_table = lpddr5_wm_table; + else + dcn42_bw_params.wm_table = ddr5_wm_table; + dcn42_bw_params.vram_type = ctx->dc_bios->integrated_info->memory_type; + dcn42_bw_params.dram_channel_width_bytes = ctx->dc_bios->integrated_info->memory_type == 0x22 ? 8 : 4; + dcn42_bw_params.num_channels = ctx->dc_bios->integrated_info->ma_channel_number ? ctx->dc_bios->integrated_info->ma_channel_number : 1; + clk_mgr->base.base.dprefclk_khz = dcn42_smu_get_dprefclk(&clk_mgr->base); + clk_mgr->base.base.clks.ref_dtbclk_khz = dcn42_smu_get_dtbclk(&clk_mgr->base); + } + /* in case we don't get a value from the BIOS, use default */ + if (clk_mgr->base.base.dentist_vco_freq_khz == 0) + clk_mgr->base.base.dentist_vco_freq_khz = 3000000; /* 3000MHz */ - /* Saved clocks configured at boot for debug purposes */ - dcn42_dump_clk_registers(&clk_mgr->base.base.boot_snapshot, clk_mgr); + /* Saved clocks configured at boot for debug purposes */ + dcn42_dump_clk_registers(&clk_mgr->base.base.boot_snapshot, clk_mgr); - if (clk_mgr->base.smu_present) - clk_mgr->base.base.dprefclk_khz = dcn42_smu_get_dprefclk(&clk_mgr->base); - clk_mgr->base.base.clks.ref_dtbclk_khz = 600000; dce_clock_read_ss_info(&clk_mgr->base); /*when clk src is from FCH, it could have ss, same clock src as DPREF clk*/ diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.h b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.h index 99fcdb602c62..5ad027a9edaf 100644 --- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.h +++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn42/dcn42_clk_mgr.h @@ -59,7 +59,6 @@ void dcn42_clk_mgr_construct(struct dc_context *ctx, void dcn42_clk_mgr_destroy(struct clk_mgr_internal *clk_mgr_int); -/* Exposed for dcn42b reuse */ void dcn42_init_single_clock(unsigned int *entry_0, uint32_t *smu_entry_0, uint8_t num_levels); @@ -76,4 +75,5 @@ int dcn42_get_active_display_cnt_wa(struct dc *dc, struct dc_state *context, int void dcn42_update_clocks_update_dpp_dto(struct clk_mgr_internal *clk_mgr, struct dc_state *context, bool safe_to_lower); void dcn42_update_clocks_update_dtb_dto(struct clk_mgr_internal *clk_mgr, struct dc_state *context, int ref_dtbclk_khz); bool dcn42_is_spll_ssc_enabled(struct clk_mgr *clk_mgr_base); +bool dcn42_has_active_display(struct dc *dc, const struct dc_state *context); #endif //__DCN42_CLK_MGR_H__ diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index 6fbdfb34379d..31589f22aae0 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -2895,16 +2895,27 @@ static struct surface_update_descriptor det_surface_update( elevate_update_type(&overall_type, UPDATE_TYPE_FAST, LOCK_DESCRIPTOR_STREAM); } - if (u->blend_tf || (u->gamma && dce_use_lut(u->plane_info ? u->plane_info->format : u->surface->format))) { + if (u->cm || (u->gamma && dce_use_lut(u->plane_info ? u->plane_info->format : u->surface->format))) { update_flags->bits.gamma_change = 1; elevate_update_type(&overall_type, UPDATE_TYPE_FAST, LOCK_DESCRIPTOR_STREAM); } - if (u->lut3d_func || u->func_shaper) { + if (u->cm && (u->cm->flags.bits.lut3d_enable || u->surface->cm.flags.bits.lut3d_enable)) { update_flags->bits.lut_3d = 1; elevate_update_type(&overall_type, UPDATE_TYPE_FAST, LOCK_DESCRIPTOR_STREAM); } + if (u->cm && u->cm->flags.bits.lut3d_dma_enable != u->surface->cm.flags.bits.lut3d_dma_enable && + u->cm->flags.bits.lut3d_enable && u->surface->cm.flags.bits.lut3d_enable) { + /* Toggling 3DLUT loading between DMA and Host is illegal */ + BREAK_TO_DEBUGGER(); + } + + if (u->cm && u->cm->flags.bits.lut3d_enable && !u->cm->flags.bits.lut3d_dma_enable) { + /* Host loading 3DLUT requires full update but only stream lock */ + elevate_update_type(&overall_type, UPDATE_TYPE_FULL, LOCK_DESCRIPTOR_STREAM); + } + if (u->hdr_mult.value) if (u->hdr_mult.value != u->surface->hdr_mult.value) { // TODO: Should be fast? @@ -2919,24 +2930,15 @@ static struct surface_update_descriptor det_surface_update( elevate_update_type(&overall_type, UPDATE_TYPE_FULL, LOCK_DESCRIPTOR_GLOBAL); } - if (u->cm2_params) { - if (u->cm2_params->component_settings.shaper_3dlut_setting != u->surface->mcm_shaper_3dlut_setting - || u->cm2_params->component_settings.lut1d_enable != u->surface->mcm_lut1d_enable - || u->cm2_params->cm2_luts.lut3d_data.lut3d_src != u->surface->mcm_luts.lut3d_data.lut3d_src) { - update_flags->bits.mcm_transfer_function_enable_change = 1; - elevate_update_type(&overall_type, UPDATE_TYPE_FULL, LOCK_DESCRIPTOR_GLOBAL); - } - } - - if (update_flags->bits.lut_3d && - u->surface->mcm_luts.lut3d_data.lut3d_src != DC_CM2_TRANSFER_FUNC_SOURCE_VIDMEM) { - elevate_update_type(&overall_type, UPDATE_TYPE_FULL, LOCK_DESCRIPTOR_GLOBAL); + if (u->cm_hist_control) { + update_flags->bits.cm_hist_change = 1; + elevate_update_type(&overall_type, UPDATE_TYPE_FAST, LOCK_DESCRIPTOR_STREAM); } - if (check_config->enable_legacy_fast_update && (update_flags->bits.gamma_change || update_flags->bits.gamut_remap_change || update_flags->bits.input_csc_change || + update_flags->bits.cm_hist_change || update_flags->bits.coeff_reduction_change)) { elevate_update_type(&overall_type, UPDATE_TYPE_FULL, LOCK_DESCRIPTOR_GLOBAL); } @@ -3168,6 +3170,11 @@ static void copy_surface_update_to_plane( surface->gamma_correction.type = srf_update->gamma->type; } + if (srf_update->cm_hist_control) { + memcpy(&surface->cm_hist_control, + srf_update->cm_hist_control, + sizeof(surface->cm_hist_control)); + } if (srf_update->in_transfer_func) { surface->in_transfer_func.sdr_ref_white_level = @@ -3181,24 +3188,12 @@ static void copy_surface_update_to_plane( sizeof(struct dc_transfer_func_distributed_points)); } - if (srf_update->cm2_params) { - surface->mcm_shaper_3dlut_setting = srf_update->cm2_params->component_settings.shaper_3dlut_setting; - surface->mcm_lut1d_enable = srf_update->cm2_params->component_settings.lut1d_enable; - surface->mcm_luts = srf_update->cm2_params->cm2_luts; - } - - if (srf_update->func_shaper) { - memcpy(&surface->in_shaper_func, srf_update->func_shaper, - sizeof(surface->in_shaper_func)); - - if (surface->mcm_shaper_3dlut_setting >= DC_CM2_SHAPER_3DLUT_SETTING_ENABLE_SHAPER) - surface->mcm_luts.shaper = &surface->in_shaper_func; + /* Shaper, 3DLUT, 1DLUT */ + if (srf_update->cm) { + memcpy(&surface->cm, srf_update->cm, + sizeof(surface->cm)); } - if (srf_update->lut3d_func) - memcpy(&surface->lut3d_func, srf_update->lut3d_func, - sizeof(surface->lut3d_func)); - if (srf_update->hdr_mult.value) surface->hdr_mult = srf_update->hdr_mult; @@ -3207,17 +3202,6 @@ static void copy_surface_update_to_plane( surface->sdr_white_level_nits = srf_update->sdr_white_level_nits; - if (srf_update->blend_tf) { - memcpy(&surface->blend_tf, srf_update->blend_tf, - sizeof(surface->blend_tf)); - - if (surface->mcm_lut1d_enable) - surface->mcm_luts.lut1d_func = &surface->blend_tf; - } - - if (srf_update->cm2_params || srf_update->blend_tf) - surface->lut_bank_a = !surface->lut_bank_a; - if (srf_update->input_csc_color_matrix) surface->input_csc_color_matrix = *srf_update->input_csc_color_matrix; @@ -4501,11 +4485,9 @@ static void commit_planes_for_stream(struct dc *dc, if (!should_update_pipe_for_plane(context, pipe_ctx, plane_state)) continue; - if (srf_updates[i].cm2_params && - srf_updates[i].cm2_params->cm2_luts.lut3d_data.lut3d_src == - DC_CM2_TRANSFER_FUNC_SOURCE_VIDMEM && - srf_updates[i].cm2_params->component_settings.shaper_3dlut_setting == - DC_CM2_SHAPER_3DLUT_SETTING_ENABLE_SHAPER_3DLUT && + if (srf_updates[i].cm && + srf_updates[i].cm->flags.bits.lut3d_enable && + srf_updates[i].cm->flags.bits.lut3d_dma_enable && dc->hwss.trigger_3dlut_dma_load) dc->hwss.trigger_3dlut_dma_load(dc, pipe_ctx); @@ -5073,6 +5055,7 @@ void populate_fast_updates(struct dc_fast_update *fast_update, fast_update[i].input_csc_color_matrix = srf_updates[i].input_csc_color_matrix; fast_update[i].coeff_reduction_factor = srf_updates[i].coeff_reduction_factor; fast_update[i].cursor_csc_color_matrix = srf_updates[i].cursor_csc_color_matrix; + fast_update[i].cm_hist_control = srf_updates[i].cm_hist_control; } } @@ -5090,6 +5073,7 @@ static bool fast_updates_exist(const struct dc_fast_update *fast_update, int sur fast_update[i].gamut_remap_matrix || fast_update[i].input_csc_color_matrix || fast_update[i].cursor_csc_color_matrix || + fast_update[i].cm_hist_control || fast_update[i].coeff_reduction_factor) return true; } @@ -5110,6 +5094,7 @@ bool fast_nonaddr_updates_exist(struct dc_fast_update *fast_update, int surface_ fast_update[i].gamma || fast_update[i].gamut_remap_matrix || fast_update[i].coeff_reduction_factor || + fast_update[i].cm_hist_control || fast_update[i].cursor_csc_color_matrix) return true; } @@ -5151,6 +5136,12 @@ static bool full_update_required( const struct dc_stream_update *stream_update, const struct dc_stream_state *stream) { + const union dc_plane_cm_flags blend_only_flags = { + .bits = { + .blend_enable = 1, + } + }; + if (full_update_required_weak(dc, srf_updates, surface_count, stream_update, stream)) return true; @@ -5163,14 +5154,12 @@ static bool full_update_required( (srf_updates[i].sdr_white_level_nits && srf_updates[i].sdr_white_level_nits != srf_updates->surface->sdr_white_level_nits) || srf_updates[i].in_transfer_func || - srf_updates[i].func_shaper || - srf_updates[i].lut3d_func || srf_updates[i].surface->force_full_update || (srf_updates[i].flip_addr && srf_updates[i].flip_addr->address.tmz_surface != srf_updates[i].surface->address.tmz_surface) || - (srf_updates[i].cm2_params && - (srf_updates[i].cm2_params->component_settings.shaper_3dlut_setting != srf_updates[i].surface->mcm_shaper_3dlut_setting || - srf_updates[i].cm2_params->component_settings.lut1d_enable != srf_updates[i].surface->mcm_lut1d_enable)))) + (srf_updates[i].cm && + ((srf_updates[i].cm->flags.all != blend_only_flags.all && srf_updates[i].cm->flags.all != 0) || + (srf_updates[i].surface->cm.flags.all != blend_only_flags.all && srf_updates[i].surface->cm.flags.all != 0))))) return true; } @@ -6898,7 +6887,7 @@ bool dc_capture_register_software_state(struct dc *dc, struct dc_register_softwa struct dc_plane_state *plane_state = pipe_ctx->plane_state; /* MPCC blending tree and mode control - capture actual blend configuration */ - state->mpc.mpcc_mode[i] = (plane_state->blend_tf.type != TF_TYPE_BYPASS) ? 1 : 0; + state->mpc.mpcc_mode[i] = (plane_state->cm.blend_func.type != TF_TYPE_BYPASS) ? 1 : 0; state->mpc.mpcc_alpha_blend_mode[i] = plane_state->per_pixel_alpha ? 1 : 0; state->mpc.mpcc_alpha_multiplied_mode[i] = plane_state->pre_multiplied_alpha ? 1 : 0; state->mpc.mpcc_blnd_active_overlap_only[i] = 0; /* Default - no overlap restriction */ diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_stream.c b/drivers/gpu/drm/amd/display/dc/core/dc_stream.c index baf820e6eae8..908f79b02102 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_stream.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_stream.c @@ -33,6 +33,7 @@ #include "dc_dmub_srv.h" #include "dc_state_priv.h" #include "dc_stream_priv.h" +#include "dce/dmub_hw_lock_mgr.h" #define DC_LOGGER dc->ctx->logger #ifndef MIN @@ -171,10 +172,12 @@ struct dc_stream_state *dc_create_stream_for_sink( goto fail; stream = kzalloc_obj(struct dc_stream_state, GFP_ATOMIC); + if (stream == NULL) goto fail; stream->update_scratch = kzalloc((int32_t) dc_update_scratch_space_size(), GFP_ATOMIC); + if (stream->update_scratch == NULL) goto fail; @@ -245,7 +248,6 @@ const struct dc_stream_status *dc_stream_get_status_const( const struct dc_stream_state *stream) { struct dc *dc = stream->ctx->dc; - return dc_state_get_stream_status(dc->current_state, stream); } @@ -257,6 +259,7 @@ void program_cursor_attributes( struct resource_context *res_ctx; struct pipe_ctx *pipe_to_program = NULL; bool enable_cursor_offload = dc_dmub_srv_is_cursor_offload_enabled(dc); + bool unlock_dmub = false; if (!stream) return; @@ -275,6 +278,12 @@ void program_cursor_attributes( if (enable_cursor_offload && dc->hwss.begin_cursor_offload_update) { dc->hwss.begin_cursor_offload_update(dc, pipe_ctx); } else { + if (dc->hwss.dmub_hw_control_lock && pipe_ctx->stream && + should_use_dmub_inbox0_lock_for_link(dc, pipe_ctx->stream->link)) { + dc->hwss.dmub_hw_control_lock(dc, dc->current_state, true); + unlock_dmub = true; + } + dc->hwss.cursor_lock(dc, pipe_to_program, true); if (pipe_to_program->next_odm_pipe) dc->hwss.cursor_lock(dc, pipe_to_program->next_odm_pipe, true); @@ -297,6 +306,9 @@ void program_cursor_attributes( dc->hwss.cursor_lock(dc, pipe_to_program, false); if (pipe_to_program->next_odm_pipe) dc->hwss.cursor_lock(dc, pipe_to_program->next_odm_pipe, false); + + if (unlock_dmub) + dc->hwss.dmub_hw_control_lock(dc, dc->current_state, false); } } } @@ -404,6 +416,7 @@ void program_cursor_position( struct resource_context *res_ctx; struct pipe_ctx *pipe_to_program = NULL; bool enable_cursor_offload = dc_dmub_srv_is_cursor_offload_enabled(dc); + bool unlock_dmub = false; if (!stream) return; @@ -423,10 +436,16 @@ void program_cursor_position( if (!pipe_to_program) { pipe_to_program = pipe_ctx; - if (enable_cursor_offload && dc->hwss.begin_cursor_offload_update) + if (enable_cursor_offload && dc->hwss.begin_cursor_offload_update) { dc->hwss.begin_cursor_offload_update(dc, pipe_ctx); - else + } else { + if (dc->hwss.dmub_hw_control_lock && pipe_ctx->stream && + should_use_dmub_inbox0_lock_for_link(dc, pipe_ctx->stream->link)) { + dc->hwss.dmub_hw_control_lock(dc, dc->current_state, true); + unlock_dmub = true; + } dc->hwss.cursor_lock(dc, pipe_to_program, true); + } } dc->hwss.set_cursor_position(pipe_ctx); @@ -438,10 +457,14 @@ void program_cursor_position( } if (pipe_to_program) { - if (enable_cursor_offload && dc->hwss.commit_cursor_offload_update) + if (enable_cursor_offload && dc->hwss.commit_cursor_offload_update) { dc->hwss.commit_cursor_offload_update(dc, pipe_to_program); - else + } else { dc->hwss.cursor_lock(dc, pipe_to_program, false); + + if (unlock_dmub) + dc->hwss.dmub_hw_control_lock(dc, dc->current_state, false); + } } } @@ -523,8 +546,10 @@ bool dc_stream_program_cursor_position( struct pipe_ctx *pipe_ctx = &dc->current_state->res_ctx.pipe_ctx[i]; /* trigger event on first pipe with current stream */ - if (stream == pipe_ctx->stream) { - pipe_ctx->stream_res.tg->funcs->program_manual_trigger(pipe_ctx->stream_res.tg); + if (stream == pipe_ctx->stream && + pipe_ctx->stream_res.tg->funcs->program_manual_trigger) { + pipe_ctx->stream_res.tg->funcs->program_manual_trigger( + pipe_ctx->stream_res.tg); break; } } @@ -984,7 +1009,6 @@ void dc_stream_release_3dlut_for_stream( if (rmcm_3dlut) { rmcm_3dlut->isInUse = false; rmcm_3dlut->stream = NULL; - rmcm_3dlut->protection_bits = 0; } } @@ -996,7 +1020,6 @@ void dc_stream_init_rmcm_3dlut(struct dc *dc) for (int i = 0; i < num_rmcm; i++) { dc->res_pool->rmcm_3dlut[i].isInUse = false; dc->res_pool->rmcm_3dlut[i].stream = NULL; - dc->res_pool->rmcm_3dlut[i].protection_bits = 0; } } diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_surface.c b/drivers/gpu/drm/amd/display/dc/core/dc_surface.c index d4c40b44d909..5f12dcca7f71 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_surface.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_surface.c @@ -139,6 +139,9 @@ const struct dc_plane_status *dc_plane_get_status( if (pipe_ctx->plane_state && flags.bits.address) pipe_ctx->plane_state->status.is_flip_pending = false; + if (pipe_ctx->plane_state && flags.bits.histogram) + memset(&pipe_ctx->plane_state->status.cm_hist, 0, + sizeof(pipe_ctx->plane_state->status.cm_hist)); break; } @@ -154,6 +157,12 @@ const struct dc_plane_status *dc_plane_get_status( if (flags.bits.address) dc->hwss.update_pending_status(pipe_ctx); + if (flags.bits.histogram) { + struct dpp *dpp = pipe_ctx->plane_res.dpp; + + if (dpp && dpp->funcs->dpp_cm_hist_read) + dpp->funcs->dpp_cm_hist_read(dpp, &pipe_ctx->plane_state->status.cm_hist); + } } return plane_status; diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h index 4bdb7bb47c75..8c49f8083475 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -1404,15 +1404,50 @@ struct lut_mem_mapping { struct dc_rmcm_3dlut { bool isInUse; const struct dc_stream_state *stream; - uint8_t protection_bits; }; struct dc_3dlut { struct kref refcount; struct tetrahedral_params lut_3d; - struct fixed31_32 hdr_multiplier; union dc_3dlut_state state; }; + +/* 3DLUT DMA (Fast Load) params */ +struct dc_3dlut_dma { + struct dc_plane_address addr; + enum dc_cm_lut_swizzle swizzle; + enum dc_cm_lut_pixel_format format; + uint16_t bias; /* FP1.5.10 */ + uint16_t scale; /* FP1.5.10 */ + enum dc_cm_lut_size size; +}; + +/* color manager */ +union dc_plane_cm_flags { + unsigned int all; + struct { + unsigned int shaper_enable : 1; + unsigned int lut3d_enable : 1; + unsigned int blend_enable : 1; + /* whether legacy (lut3d_func) or DMA is valid */ + unsigned int lut3d_dma_enable : 1; + /* RMCM lut to be used instead of MCM */ + unsigned int rmcm_enable : 1; + unsigned int reserved: 27; + } bits; +}; + +struct dc_plane_cm { + struct kref refcount; + struct dc_transfer_func shaper_func; + union { + struct dc_3dlut lut3d_func; + struct dc_3dlut_dma lut3d_dma; + }; + struct dc_transfer_func blend_func; + union dc_plane_cm_flags flags; +}; + /* * This structure is filled in by dc_surface_get_status and contains * the last requested address and the currently active address so the called @@ -1490,14 +1525,18 @@ struct dc_plane_state { struct fixed31_32 hdr_mult; struct colorspace_transform gamut_remap_matrix; - // TODO: No longer used, remove - struct dc_hdr_static_metadata hdr_static_ctx; - enum dc_color_space color_space; + bool lut_bank_a; + struct dc_hdr_static_metadata hdr_static_ctx; struct dc_3dlut lut3d_func; struct dc_transfer_func in_shaper_func; struct dc_transfer_func blend_tf; + enum dc_cm2_shaper_3dlut_setting mcm_shaper_3dlut_setting; + bool mcm_lut1d_enable; + struct dc_cm2_func_luts mcm_luts; + enum mpcc_movable_cm_location mcm_location; + struct dc_plane_cm cm; struct dc_transfer_func *gamcor_tf; enum surface_pixel_format format; @@ -1534,11 +1573,6 @@ struct dc_plane_state { bool is_statically_allocated; enum chroma_cositing cositing; - enum dc_cm2_shaper_3dlut_setting mcm_shaper_3dlut_setting; - bool mcm_lut1d_enable; - struct dc_cm2_func_luts mcm_luts; - bool lut_bank_a; - enum mpcc_movable_cm_location mcm_location; struct dc_csc_transform cursor_csc_color_matrix; bool adaptive_sharpness_en; int adaptive_sharpness_policy; @@ -1884,6 +1918,7 @@ struct dc_surface_update { * change cm2_params.cm2_luts: Fast update */ const struct dc_cm2_parameters *cm2_params; + const struct dc_plane_cm *cm; const struct dc_csc_transform *cursor_csc_color_matrix; unsigned int sdr_white_level_nits; struct dc_bias_and_scale bias_and_scale; @@ -1928,6 +1963,10 @@ struct dc_3dlut *dc_create_3dlut_func(void); void dc_3dlut_func_release(struct dc_3dlut *lut); void dc_3dlut_func_retain(struct dc_3dlut *lut); +struct dc_plane_cm *dc_plane_cm_create(void); +void dc_plane_cm_release(struct dc_plane_cm *cm); +void dc_plane_cm_retain(struct dc_plane_cm *cm); + void dc_post_update_surfaces_to_stream( struct dc *dc); diff --git a/drivers/gpu/drm/amd/display/dc/dc_bios_types.h b/drivers/gpu/drm/amd/display/dc/dc_bios_types.h index 526f71616f94..6f96c5cf39fe 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_bios_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_bios_types.h @@ -102,8 +102,7 @@ struct dc_vbios_funcs { struct bp_external_encoder_control *cntl); enum bp_result (*dac_load_detection)( struct dc_bios *bios, - enum engine_id engine_id, - struct graphics_object_id ext_enc_id); + enum engine_id engine_id); enum bp_result (*transmitter_control)( struct dc_bios *bios, struct bp_transmitter_control *cntl); diff --git a/drivers/gpu/drm/amd/display/dc/dc_plane.h b/drivers/gpu/drm/amd/display/dc/dc_plane.h index 14feb843e694..2ad6d9318566 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_plane.h +++ b/drivers/gpu/drm/amd/display/dc/dc_plane.h @@ -31,6 +31,7 @@ union dc_plane_status_update_flags { struct { uint32_t address : 1; + uint32_t histogram : 1; } bits; uint32_t raw; }; diff --git a/drivers/gpu/drm/amd/display/dc/dc_types.h b/drivers/gpu/drm/amd/display/dc/dc_types.h index d0ae5c27a511..e224077c8902 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_types.h @@ -1483,4 +1483,28 @@ struct dc_validation_dpia_set { uint32_t required_bw; }; +enum dc_cm_lut_swizzle { + CM_LUT_3D_SWIZZLE_LINEAR_RGB, + CM_LUT_3D_SWIZZLE_LINEAR_BGR, + CM_LUT_1D_PACKED_LINEAR +}; + +enum dc_cm_lut_pixel_format { + CM_LUT_PIXEL_FORMAT_RGBA16161616_UNORM_12MSB, + CM_LUT_PIXEL_FORMAT_BGRA16161616_UNORM_12MSB, + CM_LUT_PIXEL_FORMAT_RGBA16161616_UNORM_12LSB, + CM_LUT_PIXEL_FORMAT_BGRA16161616_UNORM_12LSB, + CM_LUT_PIXEL_FORMAT_RGBA16161616_FLOAT_FP1_5_10, + CM_LUT_PIXEL_FORMAT_BGRA16161616_FLOAT_FP1_5_10 +}; + +enum dc_cm_lut_size { + CM_LUT_SIZE_NONE, + CM_LUT_SIZE_999, + CM_LUT_SIZE_171717, + CM_LUT_SIZE_333333, + CM_LUT_SIZE_454545, + CM_LUT_SIZE_656565, +}; + #endif /* DC_TYPES_H_ */ diff --git a/drivers/gpu/drm/amd/display/dc/dccg/dcn401/dcn401_dccg.c b/drivers/gpu/drm/amd/display/dc/dccg/dcn401/dcn401_dccg.c index a37f94dec6f2..4b9a14c679d3 100644 --- a/drivers/gpu/drm/amd/display/dc/dccg/dcn401/dcn401_dccg.c +++ b/drivers/gpu/drm/amd/display/dc/dccg/dcn401/dcn401_dccg.c @@ -827,6 +827,16 @@ void dccg401_enable_symclk_se(struct dccg *dccg, uint32_t stream_enc_inst, uint3 if (dccg->ctx->dc->debug.root_clock_optimization.bits.symclk32_se) REG_UPDATE(DCCG_GATE_DISABLE_CNTL5, SYMCLKD_FE_ROOT_GATE_DISABLE, 1); break; + case 4: + if (dccg_dcn->dccg_mask->SYMCLKE_FE_ROOT_GATE_DISABLE) { + REG_UPDATE_2(SYMCLKE_CLOCK_ENABLE, + SYMCLKE_FE_EN, 1, + SYMCLKE_FE_SRC_SEL, link_enc_inst); + REG_UPDATE(DCCG_GATE_DISABLE_CNTL5, SYMCLKE_FE_ROOT_GATE_DISABLE, 1); + } + break; + default: + return; } } @@ -855,6 +865,16 @@ void dccg401_disable_symclk_se(struct dccg *dccg, uint32_t stream_enc_inst, uint SYMCLKD_FE_EN, 0, SYMCLKD_FE_SRC_SEL, 0); break; + case 4: + if (dccg_dcn->dccg_mask->SYMCLKE_FE_ROOT_GATE_DISABLE) { + REG_UPDATE(DCCG_GATE_DISABLE_CNTL5, SYMCLKE_FE_ROOT_GATE_DISABLE, 0); + REG_UPDATE_2(SYMCLKE_CLOCK_ENABLE, + SYMCLKE_FE_EN, 0, + SYMCLKE_FE_SRC_SEL, 0); + } + break; + default: + return; } } diff --git a/drivers/gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.c b/drivers/gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.c index 7116fdd4c7ec..d0ffa99f1fe0 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.c @@ -28,6 +28,14 @@ #include "dc_types.h" #include "core_types.h" +static bool dmub_hw_lock_has_inbox0_lock(const struct dc *dc) +{ + return dc->ctx && dc->ctx->dmub_srv && + dc->hwss.dmub_hw_control_lock && + dc->hwss.dmub_hw_control_lock_fast && + dc->ctx->dmub_srv->dmub->meta_info.feature_bits.bits.inbox0_lock_support; +} + void dmub_hw_lock_mgr_cmd(struct dc_dmub_srv *dmub_srv, bool lock, union dmub_hw_lock_flags *hw_locks, @@ -105,5 +113,13 @@ bool should_use_dmub_inbox1_lock(const struct dc *dc, const struct dc_link *link if (dc->ctx->dce_version >= DCN_VERSION_4_01) return false; + if (dmub_hw_lock_has_inbox0_lock(dc)) + return false; + return dmub_hw_lock_mgr_does_link_require_lock(dc, link); } + +bool should_use_dmub_inbox0_lock_for_link(const struct dc *dc, const struct dc_link *link) +{ + return dmub_hw_lock_has_inbox0_lock(dc) && dmub_hw_lock_mgr_does_link_require_lock(dc, link); +} diff --git a/drivers/gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.h b/drivers/gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.h index 9495857902ed..3e8caa930390 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.h +++ b/drivers/gpu/drm/amd/display/dc/dce/dmub_hw_lock_mgr.h @@ -67,4 +67,17 @@ bool dmub_hw_lock_mgr_does_link_require_lock(const struct dc *dc, const struct d */ bool dmub_hw_lock_mgr_does_context_require_lock(const struct dc *dc, const struct dc_state *context); +/** + * should_use_dmub_inbox0_lock_for_link() - Checks if the inbox0 interlock with DMU should be used. + * + * Is not functionally equivalent to inbox1 as DMUB will not own programming of the relevant locking + * registers. + * + * @dc: pointer to DC object + * @link: optional pointer to the link object to check for enabled link features + * + * Return: true if the inbox0 lock should be used, false otherwise + */ +bool should_use_dmub_inbox0_lock_for_link(const struct dc *dc, const struct dc_link *link); + #endif /*_DMUB_HW_LOCK_MGR_H_ */ diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_translation_helper.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_translation_helper.c index 6cc03391fff6..847fab508750 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_translation_helper.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_translation_helper.c @@ -601,36 +601,31 @@ static void populate_dml21_plane_config_from_plane_state(struct dml2_context *dm plane->composition.viewport.stationary = false; - if (plane_state->mcm_luts.lut3d_data.lut3d_src == DC_CM2_TRANSFER_FUNC_SOURCE_VIDMEM) { + if (plane_state->cm.flags.bits.lut3d_dma_enable) { plane->tdlut.setup_for_tdlut = true; - switch (plane_state->mcm_luts.lut3d_data.gpu_mem_params.layout) { - case DC_CM2_GPU_MEM_LAYOUT_3D_SWIZZLE_LINEAR_RGB: - case DC_CM2_GPU_MEM_LAYOUT_3D_SWIZZLE_LINEAR_BGR: + switch (plane_state->cm.lut3d_dma.swizzle) { + case CM_LUT_3D_SWIZZLE_LINEAR_RGB: + case CM_LUT_3D_SWIZZLE_LINEAR_BGR: plane->tdlut.tdlut_addressing_mode = dml2_tdlut_sw_linear; break; - case DC_CM2_GPU_MEM_LAYOUT_1D_PACKED_LINEAR: + case CM_LUT_1D_PACKED_LINEAR: + default: plane->tdlut.tdlut_addressing_mode = dml2_tdlut_simple_linear; break; } - switch (plane_state->mcm_luts.lut3d_data.gpu_mem_params.size) { - case DC_CM2_GPU_MEM_SIZE_171717: - plane->tdlut.tdlut_width_mode = dml2_tdlut_width_17_cube; - break; - case DC_CM2_GPU_MEM_SIZE_333333: + switch (plane_state->cm.lut3d_dma.size) { + case CM_LUT_SIZE_333333: plane->tdlut.tdlut_width_mode = dml2_tdlut_width_33_cube; break; - // handling when use case and HW support available - case DC_CM2_GPU_MEM_SIZE_454545: - case DC_CM2_GPU_MEM_SIZE_656565: - break; - case DC_CM2_GPU_MEM_SIZE_TRANSFORMED: + case CM_LUT_SIZE_171717: default: - //plane->tdlut.tdlut_width_mode = dml2_tdlut_width_flatten; // dml2_tdlut_width_flatten undefined + plane->tdlut.tdlut_width_mode = dml2_tdlut_width_17_cube; break; } } + plane->tdlut.setup_for_tdlut |= dml_ctx->config.force_tdlut_enable; plane->dynamic_meta_data.enable = false; @@ -967,5 +962,8 @@ void dml21_init_min_clocks_for_dc_state(struct dml2_context *in_ctx, struct dc_s min_clocks->phyclk_khz = in_ctx->v21.dml_init.soc_bb.clk_table.phyclk.clk_values_khz[lowest_dpm_state_index]; min_clocks->stutter_efficiency.base_efficiency = 1; min_clocks->stutter_efficiency.low_power_efficiency = 1; + min_clocks->stutter_efficiency.z8_stutter_efficiency = 1; + min_clocks->stutter_efficiency.z8_stutter_period = 100000; + min_clocks->zstate_support = DCN_ZSTATE_SUPPORT_ALLOW; } diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_utils.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_utils.c index 732b994b8864..ab7ec24268be 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_utils.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_utils.c @@ -374,6 +374,7 @@ void dml21_handle_phantom_streams_planes(const struct dc *dc, struct dc_state *c dml2_map_dc_pipes(dml_ctx, context, NULL, &dml_ctx->v21.dml_to_dc_pipe_mapping, dc->current_state); } + static unsigned int dml21_build_fams2_stream_programming_v2(const struct dc *dc, struct dc_state *context, struct dml2_context *dml_ctx) diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_wrapper.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_wrapper.c index 96c62bd6a37b..2623e917ec28 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_wrapper.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/dml21_wrapper.c @@ -9,16 +9,21 @@ #include "dml21_utils.h" #include "dml21_translation_helper.h" #include "dml2_dc_resource_mgmt.h" +#include "dc_fpu.h" + +#if !defined(DC_RUN_WITH_PREEMPTION_ENABLED) +#define DC_RUN_WITH_PREEMPTION_ENABLED(code) code +#endif // !DC_RUN_WITH_PREEMPTION_ENABLED #define INVALID -1 static bool dml21_allocate_memory(struct dml2_context **dml_ctx) { - *dml_ctx = vzalloc(sizeof(struct dml2_context)); + DC_RUN_WITH_PREEMPTION_ENABLED(*dml_ctx = vzalloc(sizeof(struct dml2_context))); if (!(*dml_ctx)) return false; - (*dml_ctx)->v21.dml_init.dml2_instance = vzalloc(sizeof(struct dml2_instance)); + DC_RUN_WITH_PREEMPTION_ENABLED((*dml_ctx)->v21.dml_init.dml2_instance = vzalloc(sizeof(struct dml2_instance))); if (!((*dml_ctx)->v21.dml_init.dml2_instance)) return false; @@ -28,7 +33,7 @@ static bool dml21_allocate_memory(struct dml2_context **dml_ctx) (*dml_ctx)->v21.mode_support.display_config = &(*dml_ctx)->v21.display_config; (*dml_ctx)->v21.mode_programming.display_config = (*dml_ctx)->v21.mode_support.display_config; - (*dml_ctx)->v21.mode_programming.programming = vzalloc(sizeof(struct dml2_display_cfg_programming)); + DC_RUN_WITH_PREEMPTION_ENABLED((*dml_ctx)->v21.mode_programming.programming = vzalloc(sizeof(struct dml2_display_cfg_programming))); if (!((*dml_ctx)->v21.mode_programming.programming)) return false; @@ -70,8 +75,9 @@ static void dml21_init(const struct dc *in_dc, struct dml2_context *dml_ctx, con bool dml21_create(const struct dc *in_dc, struct dml2_context **dml_ctx, const struct dml2_configuration_options *config) { /* Allocate memory for initializing DML21 instance */ - if (!dml21_allocate_memory(dml_ctx)) + if (!dml21_allocate_memory(dml_ctx)) { return false; + } dml21_init(in_dc, *dml_ctx, config); diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/bounding_boxes/dcn42_soc_bb.h b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/bounding_boxes/dcn42_soc_bb.h index 8cd074cfdf18..c75778ea7a2c 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/bounding_boxes/dcn42_soc_bb.h +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/bounding_boxes/dcn42_soc_bb.h @@ -1,5 +1,6 @@ /* SPDX-License-Identifier: MIT */ -/* Copyright 2026 Advanced Micro Devices, Inc. */ +// +// Copyright 2024 Advanced Micro Devices, Inc. #ifndef __DML_DML_DCN42_SOC_BB__ #define __DML_DML_DCN42_SOC_BB__ @@ -9,39 +10,39 @@ static const struct dml2_soc_qos_parameters dml_dcn42_variant_a_soc_qos_params = { .derate_table = { .system_active_urgent = { - .dram_derate_percent_pixel = 22, - .dram_derate_percent_vm = 0, - .dram_derate_percent_pixel_and_vm = 0, - .fclk_derate_percent = 76, - .dcfclk_derate_percent = 100, + .dram_derate_percent_pixel = 65, + .dram_derate_percent_vm = 30, + .dram_derate_percent_pixel_and_vm = 60, + .fclk_derate_percent = 80, + .dcfclk_derate_percent = 80, }, .system_active_average = { - .dram_derate_percent_pixel = 17, - .dram_derate_percent_vm = 0, - .dram_derate_percent_pixel_and_vm = 0, - .fclk_derate_percent = 57, - .dcfclk_derate_percent = 75, + .dram_derate_percent_pixel = 30, + .dram_derate_percent_vm = 30, + .dram_derate_percent_pixel_and_vm = 30, + .fclk_derate_percent = 60, + .dcfclk_derate_percent = 60, }, .dcn_mall_prefetch_urgent = { - .dram_derate_percent_pixel = 40, - .dram_derate_percent_vm = 0, - .dram_derate_percent_pixel_and_vm = 0, - .fclk_derate_percent = 83, - .dcfclk_derate_percent = 100, + .dram_derate_percent_pixel = 65, + .dram_derate_percent_vm = 30, + .dram_derate_percent_pixel_and_vm = 60, + .fclk_derate_percent = 80, + .dcfclk_derate_percent = 80, }, .dcn_mall_prefetch_average = { - .dram_derate_percent_pixel = 33, - .dram_derate_percent_vm = 0, - .dram_derate_percent_pixel_and_vm = 0, - .fclk_derate_percent = 62, - .dcfclk_derate_percent = 83, + .dram_derate_percent_pixel = 30, + .dram_derate_percent_vm = 30, + .dram_derate_percent_pixel_and_vm = 30, + .fclk_derate_percent = 60, + .dcfclk_derate_percent = 60, }, .system_idle_average = { - .dram_derate_percent_pixel = 70, - .dram_derate_percent_vm = 0, - .dram_derate_percent_pixel_and_vm = 0, - .fclk_derate_percent = 83, - .dcfclk_derate_percent = 100, + .dram_derate_percent_pixel = 30, + .dram_derate_percent_vm = 30, + .dram_derate_percent_pixel_and_vm = 30, + .fclk_derate_percent = 60, + .dcfclk_derate_percent = 60, }, }, .writeback = { @@ -50,80 +51,71 @@ static const struct dml2_soc_qos_parameters dml_dcn42_variant_a_soc_qos_params = .scaling_factor_mhz = 0, }, .qos_params = { - .dcn4x = { - .df_qos_response_time_fclk_cycles = 300, - .max_round_trip_to_furthest_cs_fclk_cycles = 350, - .mall_overhead_fclk_cycles = 50, - .meta_trip_adder_fclk_cycles = 36, - .average_transport_distance_fclk_cycles = 257, - .umc_urgent_ramp_latency_margin = 50, - .umc_max_latency_margin = 30, - .umc_average_latency_margin = 20, - .fabric_max_transport_latency_margin = 20, - .fabric_average_transport_latency_margin = 10, - - .per_uclk_dpm_params = { - { - .minimum_uclk_khz = 97 * 1000, - .urgent_ramp_uclk_cycles = 472, - .trip_to_memory_uclk_cycles = 827, - .meta_trip_to_memory_uclk_cycles = 827, - .maximum_latency_when_urgent_uclk_cycles = 72, - .average_latency_when_urgent_uclk_cycles = 61, - .maximum_latency_when_non_urgent_uclk_cycles = 827, - .average_latency_when_non_urgent_uclk_cycles = 118, - }, + .dcn32x = { + .loaded_round_trip_latency_fclk_cycles = 106, + .urgent_latency_us = { + .base_latency_us = 4, + .base_latency_pixel_vm_us = 4, + .base_latency_vm_us = 4, + .scaling_factor_fclk_us = 0, + .scaling_factor_mhz = 0, }, + .urgent_out_of_order_return_per_channel_pixel_and_vm_bytes = 4096, + .urgent_out_of_order_return_per_channel_pixel_only_bytes = 4096, + .urgent_out_of_order_return_per_channel_vm_only_bytes = 4096, }, }, - .qos_type = dml2_qos_param_type_dcn4x, + .qos_type = dml2_qos_param_type_dcn3, }; static const struct dml2_soc_bb dml2_socbb_dcn42 = { .clk_table = { + .wck_ratio = { + .clk_values_khz = {2}, + }, .uclk = { - .clk_values_khz = {97000}, + .clk_values_khz = {400000}, .num_clk_values = 1, }, .fclk = { - .clk_values_khz = {300000, 2500000}, - .num_clk_values = 2, + .clk_values_khz = {400000}, + .num_clk_values = 1, }, .dcfclk = { - .clk_values_khz = {200000, 1564000}, - .num_clk_values = 2, + .clk_values_khz = {200000}, + .num_clk_values = 1, }, .dispclk = { - .clk_values_khz = {100000, 1500000}, - .num_clk_values = 2, + .clk_values_khz = {1500000}, + .num_clk_values = 1, }, .dppclk = { - .clk_values_khz = {100000, 1500000}, - .num_clk_values = 2, + .clk_values_khz = {1500000}, + .num_clk_values = 1, }, .dtbclk = { - .clk_values_khz = {600000, 600000}, - .num_clk_values = 2, + .clk_values_khz = {600000}, + .num_clk_values = 1, }, .phyclk = { - .clk_values_khz = {810000, 810000}, - .num_clk_values = 2, + .clk_values_khz = {810000}, + .num_clk_values = 1, }, .socclk = { - .clk_values_khz = {300000, 1200000}, - .num_clk_values = 2, + .clk_values_khz = {600000}, + .num_clk_values = 1, }, .dscclk = { - .clk_values_khz = {500000, 500000}, - .num_clk_values = 2, + .clk_values_khz = {500000}, + .num_clk_values = 1, }, .phyclk_d18 = { - .clk_values_khz = {667000, 667000}, - .num_clk_values = 2, + .clk_values_khz = {667000}, + .num_clk_values = 1, }, .phyclk_d32 = { - .clk_values_khz = {625000, 625000}, - .num_clk_values = 2, + .clk_values_khz = {625000}, + .num_clk_values = 1, }, .dram_config = { .channel_width_bytes = 4, @@ -135,39 +127,39 @@ static const struct dml2_soc_bb dml2_socbb_dcn42 = { .qos_parameters = { .derate_table = { .system_active_urgent = { - .dram_derate_percent_pixel = 22, - .dram_derate_percent_vm = 0, - .dram_derate_percent_pixel_and_vm = 0, - .fclk_derate_percent = 76, - .dcfclk_derate_percent = 100, + .dram_derate_percent_pixel = 65, + .dram_derate_percent_vm = 30, + .dram_derate_percent_pixel_and_vm = 60, + .fclk_derate_percent = 80, + .dcfclk_derate_percent = 80, }, .system_active_average = { - .dram_derate_percent_pixel = 15, - .dram_derate_percent_vm = 0, - .dram_derate_percent_pixel_and_vm = 0, - .fclk_derate_percent = 57, - .dcfclk_derate_percent = 75, + .dram_derate_percent_pixel = 30, + .dram_derate_percent_vm = 30, + .dram_derate_percent_pixel_and_vm = 30, + .fclk_derate_percent = 60, + .dcfclk_derate_percent = 60, }, .dcn_mall_prefetch_urgent = { - .dram_derate_percent_pixel = 40, - .dram_derate_percent_vm = 0, - .dram_derate_percent_pixel_and_vm = 0, - .fclk_derate_percent = 83, - .dcfclk_derate_percent = 100, + .dram_derate_percent_pixel = 65, + .dram_derate_percent_vm = 30, + .dram_derate_percent_pixel_and_vm = 60, + .fclk_derate_percent = 80, + .dcfclk_derate_percent = 80, }, .dcn_mall_prefetch_average = { .dram_derate_percent_pixel = 30, - .dram_derate_percent_vm = 0, - .dram_derate_percent_pixel_and_vm = 0, - .fclk_derate_percent = 62, - .dcfclk_derate_percent = 83, + .dram_derate_percent_vm = 30, + .dram_derate_percent_pixel_and_vm = 30, + .fclk_derate_percent = 60, + .dcfclk_derate_percent = 60, }, .system_idle_average = { - .dram_derate_percent_pixel = 70, - .dram_derate_percent_vm = 0, - .dram_derate_percent_pixel_and_vm = 0, - .fclk_derate_percent = 83, - .dcfclk_derate_percent = 100, + .dram_derate_percent_pixel = 30, + .dram_derate_percent_vm = 30, + .dram_derate_percent_pixel_and_vm = 30, + .fclk_derate_percent = 60, + .dcfclk_derate_percent = 60, }, }, .writeback = { @@ -176,115 +168,25 @@ static const struct dml2_soc_bb dml2_socbb_dcn42 = { .scaling_factor_mhz = 0, }, .qos_params = { - .dcn4x = { - .df_qos_response_time_fclk_cycles = 300, - .max_round_trip_to_furthest_cs_fclk_cycles = 350, - .mall_overhead_fclk_cycles = 50, - .meta_trip_adder_fclk_cycles = 36, - .average_transport_distance_fclk_cycles = 260, - .umc_urgent_ramp_latency_margin = 50, - .umc_max_latency_margin = 30, - .umc_average_latency_margin = 20, - .fabric_max_transport_latency_margin = 20, - .fabric_average_transport_latency_margin = 10, - - .per_uclk_dpm_params = { - { - // State 1 - .minimum_uclk_khz = 0, - .urgent_ramp_uclk_cycles = 472, - .trip_to_memory_uclk_cycles = 827, - .meta_trip_to_memory_uclk_cycles = 827, - .maximum_latency_when_urgent_uclk_cycles = 72, - .average_latency_when_urgent_uclk_cycles = 72, - .maximum_latency_when_non_urgent_uclk_cycles = 827, - .average_latency_when_non_urgent_uclk_cycles = 117, - }, - { - // State 2 - .minimum_uclk_khz = 0, - .urgent_ramp_uclk_cycles = 546, - .trip_to_memory_uclk_cycles = 848, - .meta_trip_to_memory_uclk_cycles = 848, - .maximum_latency_when_urgent_uclk_cycles = 146, - .average_latency_when_urgent_uclk_cycles = 146, - .maximum_latency_when_non_urgent_uclk_cycles = 848, - .average_latency_when_non_urgent_uclk_cycles = 133, - }, - { - // State 3 - .minimum_uclk_khz = 0, - .urgent_ramp_uclk_cycles = 564, - .trip_to_memory_uclk_cycles = 853, - .meta_trip_to_memory_uclk_cycles = 853, - .maximum_latency_when_urgent_uclk_cycles = 164, - .average_latency_when_urgent_uclk_cycles = 164, - .maximum_latency_when_non_urgent_uclk_cycles = 853, - .average_latency_when_non_urgent_uclk_cycles = 136, - }, - { - // State 4 - .minimum_uclk_khz = 0, - .urgent_ramp_uclk_cycles = 613, - .trip_to_memory_uclk_cycles = 869, - .meta_trip_to_memory_uclk_cycles = 869, - .maximum_latency_when_urgent_uclk_cycles = 213, - .average_latency_when_urgent_uclk_cycles = 213, - .maximum_latency_when_non_urgent_uclk_cycles = 869, - .average_latency_when_non_urgent_uclk_cycles = 149, - }, - { - // State 5 - .minimum_uclk_khz = 0, - .urgent_ramp_uclk_cycles = 632, - .trip_to_memory_uclk_cycles = 874, - .meta_trip_to_memory_uclk_cycles = 874, - .maximum_latency_when_urgent_uclk_cycles = 232, - .average_latency_when_urgent_uclk_cycles = 232, - .maximum_latency_when_non_urgent_uclk_cycles = 874, - .average_latency_when_non_urgent_uclk_cycles = 153, - }, - { - // State 6 - .minimum_uclk_khz = 0, - .urgent_ramp_uclk_cycles = 665, - .trip_to_memory_uclk_cycles = 885, - .meta_trip_to_memory_uclk_cycles = 885, - .maximum_latency_when_urgent_uclk_cycles = 265, - .average_latency_when_urgent_uclk_cycles = 265, - .maximum_latency_when_non_urgent_uclk_cycles = 885, - .average_latency_when_non_urgent_uclk_cycles = 161, - }, - { - // State 7 - .minimum_uclk_khz = 0, - .urgent_ramp_uclk_cycles = 689, - .trip_to_memory_uclk_cycles = 895, - .meta_trip_to_memory_uclk_cycles = 895, - .maximum_latency_when_urgent_uclk_cycles = 289, - .average_latency_when_urgent_uclk_cycles = 289, - .maximum_latency_when_non_urgent_uclk_cycles = 895, - .average_latency_when_non_urgent_uclk_cycles = 167, - }, - { - // State 8 - .minimum_uclk_khz = 0, - .urgent_ramp_uclk_cycles = 716, - .trip_to_memory_uclk_cycles = 902, - .meta_trip_to_memory_uclk_cycles = 902, - .maximum_latency_when_urgent_uclk_cycles = 316, - .average_latency_when_urgent_uclk_cycles = 316, - .maximum_latency_when_non_urgent_uclk_cycles = 902, - .average_latency_when_non_urgent_uclk_cycles = 174, - }, + .dcn32x = { + .loaded_round_trip_latency_fclk_cycles = 106, + .urgent_latency_us = { + .base_latency_us = 4, + .base_latency_pixel_vm_us = 4, + .base_latency_vm_us = 4, + .scaling_factor_fclk_us = 0, + .scaling_factor_mhz = 0, }, + .urgent_out_of_order_return_per_channel_pixel_and_vm_bytes = 4096, + .urgent_out_of_order_return_per_channel_pixel_only_bytes = 4096, + .urgent_out_of_order_return_per_channel_vm_only_bytes = 4096, }, }, - .qos_type = dml2_qos_param_type_dcn4x, + .qos_type = dml2_qos_param_type_dcn3, }, .power_management_parameters = { - .dram_clk_change_blackout_us = 11.72, + .dram_clk_change_blackout_us = 29, .fclk_change_blackout_us = 0, .g7_ppt_blackout_us = 0, .stutter_enter_plus_exit_latency_us = 11, @@ -293,9 +195,9 @@ static const struct dml2_soc_bb dml2_socbb_dcn42 = { .z8_stutter_exit_latency_us = 200, }, - .vmin_limit = { - .dispclk_khz = 600 * 1000, - }, + .vmin_limit = { + .dispclk_khz = 632 * 1000, + }, .dprefclk_mhz = 600, .xtalclk_mhz = 24, @@ -307,6 +209,8 @@ static const struct dml2_soc_bb dml2_socbb_dcn42 = { .return_bus_width_bytes = 64, .hostvm_min_page_size_kbytes = 4, .gpuvm_min_page_size_kbytes = 256, + .gpuvm_max_page_table_levels = 1, + .hostvm_max_non_cached_page_table_levels = 2, .phy_downspread_percent = 0.38, .dcn_downspread_percent = 0.38, .dispclk_dppclk_vco_speed_mhz = 3000, diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/dml_top_display_cfg_types.h b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/dml_top_display_cfg_types.h index b44762e21550..4e9abe1a568d 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/dml_top_display_cfg_types.h +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/dml_top_display_cfg_types.h @@ -27,6 +27,19 @@ enum dml2_swizzle_mode { dml2_gfx11_sw_256kb_d_x, dml2_gfx11_sw_256kb_r_x, + dml2_sw_linear_256b, // GFX10 SW_LINEAR only accepts 256 byte aligned pitch + dml2_gfx10_sw_64kb_r_x, + dml2_gfx102_sw_64kb_s, + dml2_gfx102_sw_64kb_s_t, + dml2_gfx102_sw_64kb_s_x, + dml2_gfx102_sw_64kb_r_x, + + dml2_linear_64elements, // GFX7 LINEAR_ALIGNED accepts pitch alignment of the maximum of 64 elements or 256 bytes + dml2_gfx7_1d_thin, + dml2_gfx7_2d_thin_gen_zero, + dml2_gfx7_2d_thin_gen_one, + dml2_gfx7_2d_thin_arlene, + dml2_gfx7_2d_thin_anubis }; enum dml2_source_format_class { diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/dml_top_types.h b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/dml_top_types.h index c854515b49a3..98b26116cdc1 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/dml_top_types.h +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/inc/dml_top_types.h @@ -19,6 +19,7 @@ enum dml2_project_id { dml2_project_dcn4x_stage1, dml2_project_dcn4x_stage2, dml2_project_dcn4x_stage2_auto_drr_svp, + dml2_project_dcn40, dml2_project_dcn42, }; @@ -80,6 +81,7 @@ struct dml2_options { struct dml2_pmo_options pmo_options; }; + struct dml2_initialize_instance_in_out { struct dml2_instance *dml2_instance; struct dml2_options options; diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_factory.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_factory.c index 67e307fa4310..6cad99c21139 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_factory.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_factory.c @@ -21,6 +21,7 @@ bool dml2_core_create(enum dml2_project_id project_id, struct dml2_core_instance case dml2_project_dcn4x_stage1: result = false; break; + case dml2_project_dcn40: case dml2_project_dcn4x_stage2: case dml2_project_dcn4x_stage2_auto_drr_svp: out->initialize = &core_dcn4_initialize; diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_utils.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_utils.c index b57d0f6ea6a1..6930ba7ce5b7 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_utils.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_utils.c @@ -428,6 +428,9 @@ bool dml2_core_utils_is_phantom_pipe(const struct dml2_plane_parameters *plane_c unsigned int dml2_core_utils_get_tile_block_size_bytes(enum dml2_swizzle_mode sw_mode, unsigned int byte_per_pixel) { + if (dml2_core_utils_get_gfx_version(sw_mode) == 10 || dml2_core_utils_get_gfx_version(sw_mode) == 7) { + return dml2_core_utils_get_tile_block_size_bytes_backcompat(sw_mode, byte_per_pixel); + } if (sw_mode == dml2_sw_linear) return 256; @@ -459,14 +462,56 @@ unsigned int dml2_core_utils_get_tile_block_size_bytes(enum dml2_swizzle_mode sw }; } +unsigned int dml2_core_utils_get_tile_block_size_bytes_backcompat(enum dml2_swizzle_mode sw_mode, unsigned int byte_per_pixel) +{ + if (sw_mode == dml2_sw_linear_256b) + return 256; + else if (sw_mode == dml2_gfx10_sw_64kb_r_x) + return 65536; + else if (sw_mode == dml2_gfx102_sw_64kb_s) + return 65536; + else if (sw_mode == dml2_gfx102_sw_64kb_s_t) + return 65536; + else if (sw_mode == dml2_gfx102_sw_64kb_s_x) + return 65536; + else if (sw_mode == dml2_gfx102_sw_64kb_r_x) + return 65536; + else if (sw_mode == dml2_linear_64elements) + return 256; + else if (sw_mode == dml2_gfx7_1d_thin) + return 256; + else if (sw_mode == dml2_gfx7_2d_thin_gen_zero) + return (128 * 64 * byte_per_pixel); + else if (sw_mode == dml2_gfx7_2d_thin_gen_one) + return (128 * 128 * byte_per_pixel); + else if (sw_mode == dml2_gfx7_2d_thin_arlene) + return (64 * 32 * byte_per_pixel); + else if (sw_mode == dml2_gfx7_2d_thin_anubis) + return (128 * 128 * byte_per_pixel); + else { + DML_ASSERT(0); + return 256; + }; +} + bool dml2_core_utils_get_segment_horizontal_contiguous(enum dml2_swizzle_mode sw_mode, unsigned int byte_per_pixel) { - return (byte_per_pixel != 2); + if (dml2_core_utils_get_gfx_version(sw_mode) == 10 || dml2_core_utils_get_gfx_version(sw_mode) == 7) { + return dml2_core_utils_get_segment_horizontal_contiguous_backcompat(sw_mode, byte_per_pixel); + } else { + return (byte_per_pixel != 2); + } +} + +bool dml2_core_utils_get_segment_horizontal_contiguous_backcompat(enum dml2_swizzle_mode sw_mode, unsigned int byte_per_pixel) +{ + return !((byte_per_pixel == 4) && + ((sw_mode == dml2_gfx10_sw_64kb_r_x) || (sw_mode == dml2_gfx102_sw_64kb_s) || (sw_mode == dml2_gfx102_sw_64kb_s_t) || (sw_mode == dml2_gfx102_sw_64kb_s_x))); } bool dml2_core_utils_is_linear(enum dml2_swizzle_mode sw_mode) { - return sw_mode == dml2_sw_linear; + return (sw_mode == dml2_sw_linear || sw_mode == dml2_sw_linear_256b || sw_mode == dml2_linear_64elements); }; @@ -499,6 +544,20 @@ int unsigned dml2_core_utils_get_gfx_version(enum dml2_swizzle_mode sw_mode) sw_mode == dml2_gfx11_sw_256kb_d_x || sw_mode == dml2_gfx11_sw_256kb_r_x) version = 11; + else if (sw_mode == dml2_sw_linear_256b || + sw_mode == dml2_gfx10_sw_64kb_r_x || + sw_mode == dml2_gfx102_sw_64kb_s || + sw_mode == dml2_gfx102_sw_64kb_s_t || + sw_mode == dml2_gfx102_sw_64kb_s_x || + sw_mode == dml2_gfx102_sw_64kb_r_x) + version = 10; + else if (sw_mode == dml2_linear_64elements || + sw_mode == dml2_gfx7_1d_thin || + sw_mode == dml2_gfx7_2d_thin_gen_zero || + sw_mode == dml2_gfx7_2d_thin_gen_one || + sw_mode == dml2_gfx7_2d_thin_arlene || + sw_mode == dml2_gfx7_2d_thin_anubis) + version = 7; else { DML_LOG_VERBOSE("ERROR: Invalid sw_mode setting! val=%u\n", sw_mode); DML_ASSERT(0); diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_utils.h b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_utils.h index 95f0d017add4..471e73ed671c 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_utils.h +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_core/dml2_core_utils.h @@ -22,6 +22,8 @@ void dml2_core_utils_pipe_plane_mapping(const struct core_display_cfg_support_in bool dml2_core_utils_is_phantom_pipe(const struct dml2_plane_parameters *plane_cfg); unsigned int dml2_core_utils_get_tile_block_size_bytes(enum dml2_swizzle_mode sw_mode, unsigned int byte_per_pixel); bool dml2_core_utils_get_segment_horizontal_contiguous(enum dml2_swizzle_mode sw_mode, unsigned int byte_per_pixel); +unsigned int dml2_core_utils_get_tile_block_size_bytes_backcompat(enum dml2_swizzle_mode sw_mode, unsigned int byte_per_pixel); +bool dml2_core_utils_get_segment_horizontal_contiguous_backcompat(enum dml2_swizzle_mode sw_mode, unsigned int byte_per_pixel); bool dml2_core_utils_is_vertical_rotation(enum dml2_rotation_angle Scan); bool dml2_core_utils_is_linear(enum dml2_swizzle_mode sw_mode); int unsigned dml2_core_utils_get_gfx_version(enum dml2_swizzle_mode sw_mode); diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.c index 037272d2c2e6..d17e59d684fd 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_dcn4.c @@ -802,7 +802,6 @@ bool dpmm_dcn4_map_watermarks(struct dml2_dpmm_map_watermarks_params_in_out *in_ return true; } - bool dpmm_dcn42_map_watermarks(struct dml2_dpmm_map_watermarks_params_in_out *in_out) { const struct dml2_display_cfg *display_cfg = &in_out->display_cfg->display_config; diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_factory.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_factory.c index 80d4932bf154..1f2d9e97f5fd 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_factory.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_dpmm/dml2_dpmm_factory.c @@ -31,6 +31,7 @@ bool dml2_dpmm_create(enum dml2_project_id project_id, struct dml2_dpmm_instance out->map_watermarks = &dummy_map_watermarks; result = true; break; + case dml2_project_dcn40: case dml2_project_dcn4x_stage2: out->map_mode_to_soc_dpm = &dpmm_dcn3_map_mode_to_soc_dpm; out->map_watermarks = &dummy_map_watermarks; diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn4.h b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn4.h index f54fde8fba90..02da6f45cbf7 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn4.h +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn4.h @@ -10,4 +10,4 @@ bool mcg_dcn4_build_min_clock_table(struct dml2_mcg_build_min_clock_table_params_in_out *in_out); bool mcg_dcn4_unit_test(void); -#endif \ No newline at end of file +#endif diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn42.h b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn42.h index ad8a847a0a17..d4ea49e3e674 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn42.h +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_dcn42.h @@ -1,5 +1,6 @@ /* SPDX-License-Identifier: MIT */ -/* Copyright 2026 Advanced Micro Devices, Inc. */ +// +// Copyright 2026 Advanced Micro Devices, Inc. #ifndef __DML2_MCG_DCN42_H__ #define __DML2_MCG_DCN42_H__ diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_factory.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_factory.c index 1cff62b63e72..3dcd2c250633 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_factory.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_mcg/dml2_mcg_factory.c @@ -26,6 +26,7 @@ bool dml2_mcg_create(enum dml2_project_id project_id, struct dml2_mcg_instance * out->build_min_clock_table = &dummy_build_min_clock_table; result = true; break; + case dml2_project_dcn40: case dml2_project_dcn4x_stage2: case dml2_project_dcn4x_stage2_auto_drr_svp: out->build_min_clock_table = &mcg_dcn4_build_min_clock_table; diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_pmo/dml2_pmo_dcn4_fams2.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_pmo/dml2_pmo_dcn4_fams2.c index c26e100fcaf2..e8691983c0eb 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_pmo/dml2_pmo_dcn4_fams2.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_pmo/dml2_pmo_dcn4_fams2.c @@ -23,6 +23,7 @@ static const struct dml2_pmo_pstate_strategy base_strategy_list_1_display[] = { .allow_state_increase = true, }, + // Then VBlank { .per_stream_pstate_method = { dml2_pstate_method_vblank, dml2_pstate_method_na, dml2_pstate_method_na, dml2_pstate_method_na }, @@ -53,6 +54,7 @@ static const struct dml2_pmo_pstate_strategy base_strategy_list_2_display[] = { .allow_state_increase = true, }, + // Then VActive + VBlank { .per_stream_pstate_method = { dml2_pstate_method_vactive, dml2_pstate_method_vblank, dml2_pstate_method_na, dml2_pstate_method_na }, @@ -113,6 +115,7 @@ static const struct dml2_pmo_pstate_strategy base_strategy_list_3_display[] = { .allow_state_increase = true, }, + // VActive + 1 VBlank { .per_stream_pstate_method = { dml2_pstate_method_vactive, dml2_pstate_method_vactive, dml2_pstate_method_vblank, dml2_pstate_method_na }, @@ -149,6 +152,7 @@ static const struct dml2_pmo_pstate_strategy base_strategy_list_4_display[] = { .allow_state_increase = true, }, + // VActive + 1 VBlank { .per_stream_pstate_method = { dml2_pstate_method_vactive, dml2_pstate_method_vactive, dml2_pstate_method_vactive, dml2_pstate_method_vblank }, @@ -1651,6 +1655,7 @@ static bool validate_pstate_support_strategy_cofunctionality(struct dml2_pmo_ins if (svp_count > 0 && (pmo->options->disable_svp || !all_timings_support_svp(pmo, display_cfg, svp_stream_mask))) return false; + return is_config_schedulable(pmo, display_cfg, pstate_strategy); } @@ -1980,6 +1985,7 @@ static void reset_display_configuration(struct display_configuation_with_meta *d } } + static void setup_planes_for_drr_by_mask(struct display_configuation_with_meta *display_config, struct dml2_pmo_instance *pmo, int plane_mask) diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_pmo/dml2_pmo_factory.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_pmo/dml2_pmo_factory.c index 55d2464365d0..4d687fa86caa 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_pmo/dml2_pmo_factory.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_pmo/dml2_pmo_factory.c @@ -3,8 +3,8 @@ // Copyright 2024 Advanced Micro Devices, Inc. #include "dml2_pmo_factory.h" -#include "dml2_pmo_dcn4_fams2.h" #include "dml2_pmo_dcn3.h" +#include "dml2_pmo_dcn4_fams2.h" #include "dml2_external_lib_deps.h" static bool dummy_init_for_stutter(struct dml2_pmo_init_for_stutter_in_out *in_out) @@ -37,6 +37,7 @@ bool dml2_pmo_create(enum dml2_project_id project_id, struct dml2_pmo_instance * out->optimize_dcc_mcache = pmo_dcn4_fams2_optimize_dcc_mcache; result = true; break; + case dml2_project_dcn40: case dml2_project_dcn4x_stage2: out->initialize = pmo_dcn3_initialize; @@ -56,6 +57,7 @@ bool dml2_pmo_create(enum dml2_project_id project_id, struct dml2_pmo_instance * result = true; break; + case dml2_project_dcn42: case dml2_project_dcn4x_stage2_auto_drr_svp: out->initialize = pmo_dcn4_fams2_initialize; diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_pmo/dml2_pmo_factory.h b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_pmo/dml2_pmo_factory.h index b90f6263cd85..7218de1824cc 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_pmo/dml2_pmo_factory.h +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_pmo/dml2_pmo_factory.h @@ -10,4 +10,4 @@ bool dml2_pmo_create(enum dml2_project_id project_id, struct dml2_pmo_instance *out); -#endif \ No newline at end of file +#endif diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_top/dml2_top_interfaces.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_top/dml2_top_interfaces.c index 04860b6790df..a6c5031f69c1 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_top/dml2_top_interfaces.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_top/dml2_top_interfaces.c @@ -17,6 +17,7 @@ bool dml2_initialize_instance(struct dml2_initialize_instance_in_out *in_out) case dml2_project_dcn4x_stage1: case dml2_project_dcn4x_stage2: case dml2_project_dcn4x_stage2_auto_drr_svp: + case dml2_project_dcn40: case dml2_project_dcn42: return dml2_top_soc15_initialize_instance(in_out); case dml2_project_invalid: diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_top/dml2_top_legacy.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_top/dml2_top_legacy.c index 5e14d85821e2..0e3177fe9d27 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_top/dml2_top_legacy.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/dml2_top/dml2_top_legacy.c @@ -3,7 +3,6 @@ // Copyright 2024 Advanced Micro Devices, Inc. #include "dml2_top_legacy.h" -#include "dml2_top_soc15.h" #include "dml2_core_factory.h" #include "dml2_pmo_factory.h" #include "display_mode_core_structs.h" diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/inc/dml2_internal_shared_types.h b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/inc/dml2_internal_shared_types.h index a6bd75f30d20..d328d92240b4 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/inc/dml2_internal_shared_types.h +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml21/src/inc/dml2_internal_shared_types.h @@ -410,6 +410,7 @@ struct dml2_core_mode_support_in_out { } legacy; }; + struct dml2_core_mode_programming_in_out { /* * Inputs @@ -501,6 +502,7 @@ struct dml2_core_instance { bool (*populate_informative)(struct dml2_core_populate_informative_in_out *in_out); bool (*calculate_mcache_allocation)(struct dml2_calculate_mcache_allocation_in_out *in_out); + struct { struct dml2_core_internal_display_mode_lib mode_lib; } clean_me_up; @@ -753,6 +755,7 @@ struct dml2_pmo_instance { bool (*test_for_stutter)(struct dml2_pmo_test_for_stutter_in_out *in_out); bool (*optimize_for_stutter)(struct dml2_pmo_optimize_for_stutter_in_out *in_out); + struct dml2_pmo_init_data init_data; struct dml2_pmo_scratch scratch; }; diff --git a/drivers/gpu/drm/amd/display/dc/dml2_0/dml2_wrapper.c b/drivers/gpu/drm/amd/display/dc/dml2_0/dml2_wrapper.c index 307186eb6af0..408559d6fb2d 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2_0/dml2_wrapper.c +++ b/drivers/gpu/drm/amd/display/dc/dml2_0/dml2_wrapper.c @@ -84,8 +84,9 @@ static void dml2_init(const struct dc *in_dc, const struct dml2_configuration_op bool dml2_create(const struct dc *in_dc, const struct dml2_configuration_options *config, struct dml2_context **dml2) { // TODO : Temporarily add DCN_VERSION_3_2 for N-1 validation. Remove DCN_VERSION_3_2 after N-1 validation phase is complete. - if ((in_dc->debug.using_dml21) && (in_dc->ctx->dce_version >= DCN_VERSION_4_01)) + if ((in_dc->debug.using_dml21) && (in_dc->ctx->dce_version >= DCN_VERSION_4_01)) { return dml21_create(in_dc, dml2, config); + } // Allocate Mode Lib Ctx *dml2 = dml2_allocate_memory(); diff --git a/drivers/gpu/drm/amd/display/dc/dpp/dcn30/dcn30_dpp_cm.c b/drivers/gpu/drm/amd/display/dc/dpp/dcn30/dcn30_dpp_cm.c index 82eca0e7b7d0..3284084ca7ad 100644 --- a/drivers/gpu/drm/amd/display/dc/dpp/dcn30/dcn30_dpp_cm.c +++ b/drivers/gpu/drm/amd/display/dc/dpp/dcn30/dcn30_dpp_cm.c @@ -132,6 +132,8 @@ static void dpp3_power_on_gamcor_lut( if (dpp_base->ctx->dc->debug.enable_mem_low_power.bits.cm) { if (power_on) { REG_UPDATE(CM_MEM_PWR_CTRL, GAMCOR_MEM_PWR_FORCE, 0); + if (dpp_base->ctx->dc->caps.ips_v2_support) + REG_UPDATE(CM_MEM_PWR_CTRL, GAMCOR_MEM_PWR_DIS, 1); REG_WAIT(CM_MEM_PWR_STATUS, GAMCOR_MEM_PWR_STATE, 0, 1, 5); } else { dpp_base->ctx->dc->optimized_required = true; diff --git a/drivers/gpu/drm/amd/display/dc/dpp/dcn401/dcn401_dpp_dscl.c b/drivers/gpu/drm/amd/display/dc/dpp/dcn401/dcn401_dpp_dscl.c index a62c4733ed3b..8b6155f9122f 100644 --- a/drivers/gpu/drm/amd/display/dc/dpp/dcn401/dcn401_dpp_dscl.c +++ b/drivers/gpu/drm/amd/display/dc/dpp/dcn401/dcn401_dpp_dscl.c @@ -155,7 +155,12 @@ static void dpp401_power_on_dscl( if (dpp->tf_regs->DSCL_MEM_PWR_CTRL) { if (power_on) { REG_UPDATE(DSCL_MEM_PWR_CTRL, LUT_MEM_PWR_FORCE, 0); - REG_WAIT(DSCL_MEM_PWR_STATUS, LUT_MEM_PWR_STATE, 0, 1, 5); + if (dpp->base.ctx->dc->caps.ips_v2_support) { + /*hw default changes to LS*/ + REG_UPDATE(DSCL_MEM_PWR_CTRL, LUT_MEM_PWR_DIS, 1); + REG_WAIT(DSCL_MEM_PWR_STATUS, LUT_MEM_PWR_STATE, 0, 1, 100); + } else + REG_WAIT(DSCL_MEM_PWR_STATUS, LUT_MEM_PWR_STATE, 0, 1, 5); } else { if (dpp->base.ctx->dc->debug.enable_mem_low_power.bits.dscl) { dpp->base.ctx->dc->optimized_required = true; @@ -956,6 +961,15 @@ static void dpp401_dscl_program_isharp(struct dpp *dpp_base, *bs_coeffs_updated = false; PERF_TRACE(); + /*power on isharp_delta_mem first*/ + if (dpp_base->ctx->dc->caps.ips_v2_support) { + /*HW default is LS, need to wake up*/ + REG_UPDATE_2(ISHARP_DELTA_LUT_MEM_PWR_CTRL, + ISHARP_DELTA_LUT_MEM_PWR_FORCE, 0, + ISHARP_DELTA_LUT_MEM_PWR_DIS, 1); + REG_WAIT(ISHARP_DELTA_LUT_MEM_PWR_CTRL, + ISHARP_DELTA_LUT_MEM_PWR_STATE, 0, 1, 100); + } /* ISHARP_MODE */ REG_SET_6(ISHARP_MODE, 0, ISHARP_EN, scl_data->dscl_prog_data.isharp_en, @@ -1033,6 +1047,13 @@ static void dpp401_dscl_program_isharp(struct dpp *dpp_base, } } + /*power on isharp_delta_mem first*/ + if (dpp_base->ctx->dc->caps.ips_v2_support) { + /*HW default is LS, need to wake up*/ + REG_UPDATE_SEQ_2(ISHARP_DELTA_LUT_MEM_PWR_CTRL, + ISHARP_DELTA_LUT_MEM_PWR_FORCE, 0, + ISHARP_DELTA_LUT_MEM_PWR_DIS, 0); + } PERF_TRACE(); } // dpp401_dscl_program_isharp /** diff --git a/drivers/gpu/drm/amd/display/dc/hpo/dcn31/dcn31_hpo_dp_stream_encoder.c b/drivers/gpu/drm/amd/display/dc/hpo/dcn31/dcn31_hpo_dp_stream_encoder.c index 759b453385c4..92ed130aeaec 100644 --- a/drivers/gpu/drm/amd/display/dc/hpo/dcn31/dcn31_hpo_dp_stream_encoder.c +++ b/drivers/gpu/drm/amd/display/dc/hpo/dcn31/dcn31_hpo_dp_stream_encoder.c @@ -640,6 +640,11 @@ static void dcn31_hpo_dp_stream_enc_audio_setup( REG_UPDATE(DP_STREAM_ENC_AUDIO_CONTROL, DP_STREAM_ENC_INPUT_MUX_AUDIO_STREAM_SOURCE_SEL, az_inst); + if (enc3->hpo_se_mask->DP_STREAM_ENC_APG_CLOCK_EN) { + /*enable apg clk*/ + REG_UPDATE(DP_STREAM_ENC_AUDIO_CONTROL, + DP_STREAM_ENC_APG_CLOCK_EN, 1); + } ASSERT(enc->apg); enc->apg->funcs->se_audio_setup(enc->apg, az_inst, info); } diff --git a/drivers/gpu/drm/amd/display/dc/hubp/dcn401/dcn401_hubp.c b/drivers/gpu/drm/amd/display/dc/hubp/dcn401/dcn401_hubp.c index c205500290ec..4985e885952d 100644 --- a/drivers/gpu/drm/amd/display/dc/hubp/dcn401/dcn401_hubp.c +++ b/drivers/gpu/drm/amd/display/dc/hubp/dcn401/dcn401_hubp.c @@ -41,12 +41,12 @@ hubp2->hubp_shift->field_name, hubp2->hubp_mask->field_name void hubp401_program_3dlut_fl_addr(struct hubp *hubp, - const struct dc_plane_address address) + const struct dc_plane_address *address) { struct dcn20_hubp *hubp2 = TO_DCN20_HUBP(hubp); - REG_UPDATE(HUBP_3DLUT_ADDRESS_HIGH, HUBP_3DLUT_ADDRESS_HIGH, address.lut3d.addr.high_part); - REG_WRITE(HUBP_3DLUT_ADDRESS_LOW, address.lut3d.addr.low_part); + REG_UPDATE(HUBP_3DLUT_ADDRESS_HIGH, HUBP_3DLUT_ADDRESS_HIGH, address->lut3d.addr.high_part); + REG_WRITE(HUBP_3DLUT_ADDRESS_LOW, address->lut3d.addr.low_part); } void hubp401_program_3dlut_fl_dlg_param(struct hubp *hubp, int refcyc_per_3dlut_group) @@ -72,33 +72,46 @@ int hubp401_get_3dlut_fl_done(struct hubp *hubp) return ret; } -void hubp401_program_3dlut_fl_addressing_mode(struct hubp *hubp, enum hubp_3dlut_fl_addressing_mode addr_mode) +static void hubp401_get_3dlut_fl_xbar_map( + const enum dc_cm_lut_pixel_format format, + enum hubp_3dlut_fl_crossbar_bit_slice *bit_slice_y_g, + enum hubp_3dlut_fl_crossbar_bit_slice *bit_slice_cb_b, + enum hubp_3dlut_fl_crossbar_bit_slice *bit_slice_cr_r) { - struct dcn20_hubp *hubp2 = TO_DCN20_HUBP(hubp); - - REG_UPDATE(HUBP_3DLUT_CONTROL, HUBP_3DLUT_ADDRESSING_MODE, addr_mode); -} - -void hubp401_program_3dlut_fl_width(struct hubp *hubp, enum hubp_3dlut_fl_width width) -{ - struct dcn20_hubp *hubp2 = TO_DCN20_HUBP(hubp); - - REG_UPDATE(HUBP_3DLUT_CONTROL, HUBP_3DLUT_WIDTH, width); + switch (format) { + case CM_LUT_PIXEL_FORMAT_BGRA16161616_UNORM_12MSB: + case CM_LUT_PIXEL_FORMAT_BGRA16161616_UNORM_12LSB: + case CM_LUT_PIXEL_FORMAT_BGRA16161616_FLOAT_FP1_5_10: + /* BGRA */ + *bit_slice_cr_r = hubp_3dlut_fl_crossbar_bit_slice_32_47; + *bit_slice_y_g = hubp_3dlut_fl_crossbar_bit_slice_16_31; + *bit_slice_cb_b = hubp_3dlut_fl_crossbar_bit_slice_0_15; + break; + case CM_LUT_PIXEL_FORMAT_RGBA16161616_UNORM_12MSB: + case CM_LUT_PIXEL_FORMAT_RGBA16161616_UNORM_12LSB: + case CM_LUT_PIXEL_FORMAT_RGBA16161616_FLOAT_FP1_5_10: + default: + /* RGBA */ + *bit_slice_cr_r = hubp_3dlut_fl_crossbar_bit_slice_0_15; + *bit_slice_y_g = hubp_3dlut_fl_crossbar_bit_slice_16_31; + *bit_slice_cb_b = hubp_3dlut_fl_crossbar_bit_slice_32_47; + break; + } } -void hubp401_program_3dlut_fl_tmz_protected(struct hubp *hubp, uint8_t protection_bits) +void hubp401_program_3dlut_fl_crossbar(struct hubp *hubp, + const enum dc_cm_lut_pixel_format format) { struct dcn20_hubp *hubp2 = TO_DCN20_HUBP(hubp); - REG_UPDATE(HUBP_3DLUT_CONTROL, HUBP_3DLUT_TMZ, protection_bits); -} + enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_y_g = 0; + enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_cb_b = 0; + enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_cr_r = 0; -void hubp401_program_3dlut_fl_crossbar(struct hubp *hubp, - enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_y_g, - enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_cb_b, - enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_cr_r) -{ - struct dcn20_hubp *hubp2 = TO_DCN20_HUBP(hubp); + hubp401_get_3dlut_fl_xbar_map(format, + &bit_slice_y_g, + &bit_slice_cb_b, + &bit_slice_cr_r); REG_UPDATE_3(HUBP_3DLUT_CONTROL, HUBP_3DLUT_CROSSBAR_SELECT_Y_G, bit_slice_y_g, @@ -106,62 +119,122 @@ void hubp401_program_3dlut_fl_crossbar(struct hubp *hubp, HUBP_3DLUT_CROSSBAR_SELECT_CR_R, bit_slice_cr_r); } -void hubp401_update_3dlut_fl_bias_scale(struct hubp *hubp, uint16_t bias, uint16_t scale) +static enum hubp_3dlut_fl_width hubp401_get_3dlut_fl_width( + const enum dc_cm_lut_size size, + const enum dc_cm_lut_swizzle swizzle) { - struct dcn20_hubp *hubp2 = TO_DCN20_HUBP(hubp); + enum hubp_3dlut_fl_width width = 0; + + switch (size) { + case CM_LUT_SIZE_333333: + ASSERT(swizzle != CM_LUT_1D_PACKED_LINEAR); + width = hubp_3dlut_fl_width_33; + break; + case CM_LUT_SIZE_171717: + if (swizzle != CM_LUT_1D_PACKED_LINEAR) { + width = hubp_3dlut_fl_width_17; + } else { + width = hubp_3dlut_fl_width_17_transformed; + } + break; + default: + width = 0; + break; + } - REG_UPDATE_2(_3DLUT_FL_BIAS_SCALE, HUBP0_3DLUT_FL_BIAS, bias, HUBP0_3DLUT_FL_SCALE, scale); + return width; } -void hubp401_program_3dlut_fl_mode(struct hubp *hubp, enum hubp_3dlut_fl_mode mode) +static enum hubp_3dlut_fl_format hubp401_get_3dlut_fl_format( + const enum dc_cm_lut_pixel_format dc_format) { - struct dcn20_hubp *hubp2 = TO_DCN20_HUBP(hubp); + enum hubp_3dlut_fl_format hubp_format = hubp_3dlut_fl_format_unorm_12msb_bitslice; - REG_UPDATE(_3DLUT_FL_CONFIG, HUBP0_3DLUT_FL_MODE, mode); + switch (dc_format) { + case CM_LUT_PIXEL_FORMAT_RGBA16161616_UNORM_12MSB: + case CM_LUT_PIXEL_FORMAT_BGRA16161616_UNORM_12MSB: + hubp_format = hubp_3dlut_fl_format_unorm_12msb_bitslice; + break; + case CM_LUT_PIXEL_FORMAT_RGBA16161616_UNORM_12LSB: + case CM_LUT_PIXEL_FORMAT_BGRA16161616_UNORM_12LSB: + hubp_format = hubp_3dlut_fl_format_unorm_12lsb_bitslice; + break; + case CM_LUT_PIXEL_FORMAT_RGBA16161616_FLOAT_FP1_5_10: + case CM_LUT_PIXEL_FORMAT_BGRA16161616_FLOAT_FP1_5_10: + hubp_format = hubp_3dlut_fl_format_float_fp1_5_10; + break; + default: + BREAK_TO_DEBUGGER(); + break; + } + + return hubp_format; } -void hubp401_program_3dlut_fl_format(struct hubp *hubp, enum hubp_3dlut_fl_format format) +static enum hubp_3dlut_fl_addressing_mode hubp401_get_3dlut_fl_addr_mode( + const enum dc_cm_lut_swizzle swizzle) { - struct dcn20_hubp *hubp2 = TO_DCN20_HUBP(hubp); + enum hubp_3dlut_fl_addressing_mode addr_mode; + + switch (swizzle) { + case CM_LUT_1D_PACKED_LINEAR: + addr_mode = hubp_3dlut_fl_addressing_mode_simple_linear; + break; + case CM_LUT_3D_SWIZZLE_LINEAR_RGB: + case CM_LUT_3D_SWIZZLE_LINEAR_BGR: + default: + addr_mode = hubp_3dlut_fl_addressing_mode_sw_linear; + break; + } - REG_UPDATE(_3DLUT_FL_CONFIG, HUBP0_3DLUT_FL_FORMAT, format); + return addr_mode; } -void hubp401_program_3dlut_fl_config( - struct hubp *hubp, - struct hubp_fl_3dlut_config *cfg) +static enum hubp_3dlut_fl_mode hubp401_get_3dlut_fl_mode( + const enum dc_cm_lut_swizzle swizzle) { - struct dcn20_hubp *hubp2 = TO_DCN20_HUBP(hubp); + enum hubp_3dlut_fl_mode mode; - uint32_t mpc_width = {(cfg->width == 17) ? 0 : 1}; - uint32_t width = {cfg->width}; + switch (swizzle) { + case CM_LUT_3D_SWIZZLE_LINEAR_RGB: + mode = hubp_3dlut_fl_mode_native_1; + break; + case CM_LUT_3D_SWIZZLE_LINEAR_BGR: + mode = hubp_3dlut_fl_mode_native_2; + break; + case CM_LUT_1D_PACKED_LINEAR: + mode = hubp_3dlut_fl_mode_transform; + break; + default: + mode = hubp_3dlut_fl_mode_disable; + break; + } - if (cfg->layout == DC_CM2_GPU_MEM_LAYOUT_1D_PACKED_LINEAR) - width = (cfg->width == 17) ? 4916 : 35940; + return mode; +} + +void hubp401_program_3dlut_fl_config(struct hubp *hubp, + const struct dc_3dlut_dma *config) +{ + struct dcn20_hubp *hubp2 = TO_DCN20_HUBP(hubp); + + enum hubp_3dlut_fl_width width = hubp401_get_3dlut_fl_width(config->size, config->swizzle); + enum hubp_3dlut_fl_format format = hubp401_get_3dlut_fl_format(config->format); + enum hubp_3dlut_fl_addressing_mode addr_mode = hubp401_get_3dlut_fl_addr_mode(config->swizzle); + enum hubp_3dlut_fl_mode mode = hubp401_get_3dlut_fl_mode(config->swizzle); REG_UPDATE_2(_3DLUT_FL_CONFIG, - HUBP0_3DLUT_FL_MODE, cfg->mode, - HUBP0_3DLUT_FL_FORMAT, cfg->format); + HUBP0_3DLUT_FL_MODE, mode, + HUBP0_3DLUT_FL_FORMAT, format); REG_UPDATE_2(_3DLUT_FL_BIAS_SCALE, - HUBP0_3DLUT_FL_BIAS, cfg->bias, - HUBP0_3DLUT_FL_SCALE, cfg->scale); - - REG_UPDATE(HUBP_3DLUT_ADDRESS_HIGH, - HUBP_3DLUT_ADDRESS_HIGH, cfg->address.lut3d.addr.high_part); - REG_UPDATE(HUBP_3DLUT_ADDRESS_LOW, - HUBP_3DLUT_ADDRESS_LOW, cfg->address.lut3d.addr.low_part); - - //cross bar - REG_UPDATE_8(HUBP_3DLUT_CONTROL, - HUBP_3DLUT_MPC_WIDTH, mpc_width, - HUBP_3DLUT_WIDTH, width, - HUBP_3DLUT_CROSSBAR_SELECT_CR_R, cfg->crossbar_bit_slice_cr_r, - HUBP_3DLUT_CROSSBAR_SELECT_Y_G, cfg->crossbar_bit_slice_y_g, - HUBP_3DLUT_CROSSBAR_SELECT_CB_B, cfg->crossbar_bit_slice_cb_b, - HUBP_3DLUT_ADDRESSING_MODE, cfg->addr_mode, - HUBP_3DLUT_TMZ, cfg->protection_bits, - HUBP_3DLUT_ENABLE, cfg->enabled ? 1 : 0); + HUBP0_3DLUT_FL_BIAS, config->bias, + HUBP0_3DLUT_FL_SCALE, config->scale); + + REG_UPDATE_3(HUBP_3DLUT_CONTROL, + HUBP_3DLUT_WIDTH, width, + HUBP_3DLUT_ADDRESSING_MODE, addr_mode, + HUBP_3DLUT_TMZ, config->addr.tmz_surface); } void hubp401_update_mall_sel(struct hubp *hubp, uint32_t mall_sel, bool c_cursor) @@ -1058,19 +1131,13 @@ static struct hubp_funcs dcn401_hubp_funcs = { .hubp_update_mall_sel = hubp401_update_mall_sel, .hubp_prepare_subvp_buffering = hubp32_prepare_subvp_buffering, .hubp_program_mcache_id_and_split_coordinate = hubp401_program_mcache_id_and_split_coordinate, - .hubp_update_3dlut_fl_bias_scale = hubp401_update_3dlut_fl_bias_scale, - .hubp_program_3dlut_fl_mode = hubp401_program_3dlut_fl_mode, - .hubp_program_3dlut_fl_format = hubp401_program_3dlut_fl_format, .hubp_program_3dlut_fl_addr = hubp401_program_3dlut_fl_addr, + .hubp_program_3dlut_fl_config = hubp401_program_3dlut_fl_config, .hubp_program_3dlut_fl_dlg_param = hubp401_program_3dlut_fl_dlg_param, .hubp_enable_3dlut_fl = hubp401_enable_3dlut_fl, - .hubp_program_3dlut_fl_addressing_mode = hubp401_program_3dlut_fl_addressing_mode, - .hubp_program_3dlut_fl_width = hubp401_program_3dlut_fl_width, - .hubp_program_3dlut_fl_tmz_protected = hubp401_program_3dlut_fl_tmz_protected, .hubp_program_3dlut_fl_crossbar = hubp401_program_3dlut_fl_crossbar, .hubp_get_3dlut_fl_done = hubp401_get_3dlut_fl_done, .hubp_clear_tiling = hubp401_clear_tiling, - .hubp_program_3dlut_fl_config = hubp401_program_3dlut_fl_config, .hubp_read_reg_state = hubp3_read_reg_state }; diff --git a/drivers/gpu/drm/amd/display/dc/hubp/dcn401/dcn401_hubp.h b/drivers/gpu/drm/amd/display/dc/hubp/dcn401/dcn401_hubp.h index 4570b8016de5..043948f64b86 100644 --- a/drivers/gpu/drm/amd/display/dc/hubp/dcn401/dcn401_hubp.h +++ b/drivers/gpu/drm/amd/display/dc/hubp/dcn401/dcn401_hubp.h @@ -328,32 +328,17 @@ int hubp401_get_3dlut_fl_done(struct hubp *hubp); void hubp401_set_unbounded_requesting(struct hubp *hubp, bool enable); -void hubp401_update_3dlut_fl_bias_scale(struct hubp *hubp, uint16_t bias, uint16_t scale); - void hubp401_program_3dlut_fl_crossbar(struct hubp *hubp, - enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_y_g, - enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_cb_b, - enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_cr_r); - -void hubp401_program_3dlut_fl_tmz_protected(struct hubp *hubp, uint8_t protection_bits); - -void hubp401_program_3dlut_fl_width(struct hubp *hubp, enum hubp_3dlut_fl_width width); - -void hubp401_program_3dlut_fl_addressing_mode(struct hubp *hubp, enum hubp_3dlut_fl_addressing_mode addr_mode); + const enum dc_cm_lut_pixel_format format); void hubp401_enable_3dlut_fl(struct hubp *hubp, bool enable); void hubp401_program_3dlut_fl_dlg_param(struct hubp *hubp, int refcyc_per_3dlut_group); -void hubp401_program_3dlut_fl_addr(struct hubp *hubp, const struct dc_plane_address address); +void hubp401_program_3dlut_fl_addr(struct hubp *hubp, const struct dc_plane_address *address); -void hubp401_program_3dlut_fl_format(struct hubp *hubp, enum hubp_3dlut_fl_format format); - -void hubp401_program_3dlut_fl_mode(struct hubp *hubp, enum hubp_3dlut_fl_mode mode); - -void hubp401_program_3dlut_fl_config( - struct hubp *hubp, - struct hubp_fl_3dlut_config *cfg); +void hubp401_program_3dlut_fl_config(struct hubp *hubp, + const struct dc_3dlut_dma *config); void hubp401_clear_tiling(struct hubp *hubp); diff --git a/drivers/gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.c b/drivers/gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.c index 07c38dc03960..0e33c739f459 100644 --- a/drivers/gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.c +++ b/drivers/gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.c @@ -147,13 +147,16 @@ static void hubp42_program_pixel_format( /* don't see the need of program the xbar in DCN 1.0 */ } -void hubp42_program_deadline( +static void hubp42_program_deadline( struct hubp *hubp, struct dml2_display_dlg_regs *dlg_attr, struct dml2_display_ttu_regs *ttu_attr) { struct dcn20_hubp *hubp2 = TO_DCN20_HUBP(hubp); + /* put DLG in mission mode */ + REG_WRITE(HUBPREQ_DEBUG_DB, 0); + /* DLG - Per hubp */ REG_SET_2(BLANK_OFFSET_0, 0, REFCYC_H_BLANK_END, dlg_attr->refcyc_h_blank_end, @@ -274,19 +277,84 @@ static void hubp42_program_surface_config( hubp42_program_pixel_format(hubp, format); } +static void hubp42_get_3dlut_fl_xbar_map( + const enum dc_cm_lut_pixel_format format, + enum hubp_3dlut_fl_crossbar_bit_slice *bit_slice_y_g, + enum hubp_3dlut_fl_crossbar_bit_slice *bit_slice_cb_b, + enum hubp_3dlut_fl_crossbar_bit_slice *bit_slice_cr_r) +{ + switch (format) { + case CM_LUT_PIXEL_FORMAT_BGRA16161616_UNORM_12MSB: + case CM_LUT_PIXEL_FORMAT_BGRA16161616_UNORM_12LSB: + case CM_LUT_PIXEL_FORMAT_BGRA16161616_FLOAT_FP1_5_10: + /* BGRA */ + *bit_slice_cr_r = hubp_3dlut_fl_crossbar_bit_slice_32_47; + *bit_slice_y_g = hubp_3dlut_fl_crossbar_bit_slice_16_31; + *bit_slice_cb_b = hubp_3dlut_fl_crossbar_bit_slice_0_15; + break; + case CM_LUT_PIXEL_FORMAT_RGBA16161616_UNORM_12MSB: + case CM_LUT_PIXEL_FORMAT_RGBA16161616_UNORM_12LSB: + case CM_LUT_PIXEL_FORMAT_RGBA16161616_FLOAT_FP1_5_10: + default: + /* RGBA */ + *bit_slice_cr_r = hubp_3dlut_fl_crossbar_bit_slice_0_15; + *bit_slice_y_g = hubp_3dlut_fl_crossbar_bit_slice_16_31; + *bit_slice_cb_b = hubp_3dlut_fl_crossbar_bit_slice_32_47; + break; + } +} + void hubp42_program_3dlut_fl_crossbar(struct hubp *hubp, - enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_r, - enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_g, - enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_b) + const enum dc_cm_lut_pixel_format format) { struct dcn20_hubp *hubp2 = TO_DCN20_HUBP(hubp); + enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_g = 0; + enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_b = 0; + enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_r = 0; + + hubp42_get_3dlut_fl_xbar_map(format, + &bit_slice_g, + &bit_slice_b, + &bit_slice_r); + REG_UPDATE_3(HUBP_3DLUT_CONTROL, HUBP_3DLUT_CROSSBAR_SEL_R, bit_slice_r, HUBP_3DLUT_CROSSBAR_SEL_G, bit_slice_g, HUBP_3DLUT_CROSSBAR_SEL_B, bit_slice_b); } +static uint32_t hubp42_get_3dlut_fl_mpc_width( + const enum dc_cm_lut_size size) +{ + uint32_t width = 0; + + switch (size) { + case CM_LUT_SIZE_333333: + width = 1; + break; + case CM_LUT_SIZE_171717: + default: + width = 0; + break; + } + + return width; +} + +void hubp42_program_3dlut_fl_config(struct hubp *hubp, + const struct dc_3dlut_dma *config) +{ + struct dcn20_hubp *hubp2 = TO_DCN20_HUBP(hubp); + + uint32_t mpc_width = hubp42_get_3dlut_fl_mpc_width(config->size); + + REG_UPDATE(HUBP_3DLUT_CONTROL, + HUBP_3DLUT_MPC_WIDTH, mpc_width); + + hubp401_program_3dlut_fl_config(hubp, config); +} + static bool hubp42_program_surface_flip_and_addr( struct hubp *hubp, const struct dc_plane_address *address, @@ -548,6 +616,7 @@ struct hubp_funcs dcn42_hubp_funcs = { .hubp_setup_interdependent2 = hubp401_setup_interdependent, .hubp_set_vm_system_aperture_settings = hubp3_set_vm_system_aperture_settings, .set_blank = hubp2_set_blank, + .set_blank_regs = hubp2_set_blank_regs, .dcc_control = hubp3_dcc_control, .hubp_reset = hubp_reset, .mem_program_viewport = min_set_viewport, @@ -567,17 +636,13 @@ struct hubp_funcs dcn42_hubp_funcs = { .hubp_set_flip_int = hubp1_set_flip_int, .hubp_in_blank = hubp1_in_blank, .program_extended_blank = hubp31_program_extended_blank_value, - .hubp_update_3dlut_fl_bias_scale = hubp401_update_3dlut_fl_bias_scale, - .hubp_program_3dlut_fl_mode = hubp401_program_3dlut_fl_mode, - .hubp_program_3dlut_fl_format = hubp401_program_3dlut_fl_format, .hubp_program_3dlut_fl_addr = hubp401_program_3dlut_fl_addr, + .hubp_program_3dlut_fl_config = hubp42_program_3dlut_fl_config, .hubp_program_3dlut_fl_dlg_param = hubp401_program_3dlut_fl_dlg_param, .hubp_enable_3dlut_fl = hubp401_enable_3dlut_fl, - .hubp_program_3dlut_fl_addressing_mode = hubp401_program_3dlut_fl_addressing_mode, - .hubp_program_3dlut_fl_width = hubp401_program_3dlut_fl_width, - .hubp_program_3dlut_fl_tmz_protected = hubp401_program_3dlut_fl_tmz_protected, .hubp_program_3dlut_fl_crossbar = hubp42_program_3dlut_fl_crossbar, .hubp_get_3dlut_fl_done = hubp401_get_3dlut_fl_done, + .hubp_clear_tiling = hubp3_clear_tiling, .hubp_program_3dlut_fl_config = hubp401_program_3dlut_fl_config, .hubp_read_reg_state = hubp3_read_reg_state }; diff --git a/drivers/gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.h b/drivers/gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.h index 976614f38981..486c8907413a 100644 --- a/drivers/gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.h +++ b/drivers/gpu/drm/amd/display/dc/hubp/dcn42/dcn42_hubp.h @@ -56,11 +56,11 @@ bool hubp42_construct( const struct dcn_hubp2_shift *hubp_shift, const struct dcn_hubp2_mask *hubp_mask); -void hubp42_program_3dlut_fl_crossbar( - struct hubp *hubp, - enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_r, - enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_g, - enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_b); +void hubp42_program_3dlut_fl_crossbar(struct hubp *hubp, + const enum dc_cm_lut_pixel_format format); + +void hubp42_program_3dlut_fl_config(struct hubp *hubp, + const struct dc_3dlut_dma *config); void hubp42_read_state(struct hubp *hubp); @@ -70,10 +70,4 @@ void hubp42_setup( union dml2_global_sync_programming *pipe_global_sync, struct dc_crtc_timing *timing); -void hubp42_program_deadline( - struct hubp *hubp, - struct dml2_display_dlg_regs *dlg_attr, - struct dml2_display_ttu_regs *ttu_attr); - - #endif /* __DC_HUBP_DCN42_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c index 38a5f312ddb4..8a17cc036399 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dce110/dce110_hwseq.c @@ -661,45 +661,16 @@ void dce110_update_info_frame(struct pipe_ctx *pipe_ctx) } static void -dce110_external_encoder_control(enum bp_external_encoder_control_action action, - struct dc_link *link, - struct dc_crtc_timing *timing) +dce110_dac_encoder_control(struct pipe_ctx *pipe_ctx, bool enable) { - struct dc *dc = link->ctx->dc; + struct dc_link *link = pipe_ctx->stream->link; struct dc_bios *bios = link->ctx->dc_bios; - const struct dc_link_settings *link_settings = &link->cur_link_settings; - enum bp_result bp_result = BP_RESULT_OK; - struct bp_external_encoder_control ext_cntl = { - .action = action, - .connector_obj_id = link->link_enc->connector, - .encoder_id = link->ext_enc_id, - .lanes_number = link_settings->lane_count, - .link_rate = link_settings->link_rate, - - /* Use signal type of the real link encoder, ie. DP */ - .signal = link->connector_signal, - - /* We don't know the timing yet when executing the SETUP action, - * so use a reasonably high default value. It seems that ENABLE - * can change the actual pixel clock but doesn't work with higher - * pixel clocks than what SETUP was called with. - */ - .pixel_clock = timing ? timing->pix_clk_100hz / 10 : 300000, - .color_depth = timing ? timing->display_color_depth : COLOR_DEPTH_888, - }; - DC_LOGGER_INIT(dc->ctx); - - bp_result = bios->funcs->external_encoder_control(bios, &ext_cntl); + struct bp_encoder_control encoder_control = {0}; - if (bp_result != BP_RESULT_OK) - DC_LOG_ERROR("Failed to execute external encoder action: 0x%x\n", action); -} - -static void -dce110_prepare_ddc(struct dc_link *link) -{ - if (link->ext_enc_id.id) - dce110_external_encoder_control(EXTERNAL_ENCODER_CONTROL_DDC_SETUP, link, NULL); + encoder_control.action = enable ? ENCODER_CONTROL_ENABLE : ENCODER_CONTROL_DISABLE; + encoder_control.engine_id = link->link_enc->analog_engine; + encoder_control.pixel_clock = pipe_ctx->stream->timing.pix_clk_100hz / 10; + bios->funcs->encoder_control(bios, &encoder_control); } static bool @@ -709,8 +680,7 @@ dce110_dac_load_detect(struct dc_link *link) struct link_encoder *link_enc = link->link_enc; enum bp_result bp_result; - bp_result = bios->funcs->dac_load_detection( - bios, link_enc->analog_engine, link->ext_enc_id); + bp_result = bios->funcs->dac_load_detection(bios, link_enc->analog_engine); return bp_result == BP_RESULT_OK; } @@ -726,6 +696,7 @@ void dce110_enable_stream(struct pipe_ctx *pipe_ctx) uint32_t early_control = 0; struct timing_generator *tg = pipe_ctx->stream_res.tg; + link_hwss->setup_stream_attribute(pipe_ctx); link_hwss->setup_stream_encoder(pipe_ctx); dc->hwss.update_info_frame(pipe_ctx); @@ -744,8 +715,8 @@ void dce110_enable_stream(struct pipe_ctx *pipe_ctx) tg->funcs->set_early_control(tg, early_control); - if (link->ext_enc_id.id) - dce110_external_encoder_control(EXTERNAL_ENCODER_CONTROL_ENABLE, link, timing); + if (dc_is_rgb_signal(pipe_ctx->stream->signal)) + dce110_dac_encoder_control(pipe_ctx, true); } static enum bp_result link_transmitter_control( @@ -1244,8 +1215,8 @@ void dce110_disable_stream(struct pipe_ctx *pipe_ctx) link_enc->transmitter - TRANSMITTER_UNIPHY_A); } - if (link->ext_enc_id.id) - dce110_external_encoder_control(EXTERNAL_ENCODER_CONTROL_DISABLE, link, NULL); + if (dc_is_rgb_signal(pipe_ctx->stream->signal)) + dce110_dac_encoder_control(pipe_ctx, false); } void dce110_unblank_stream(struct pipe_ctx *pipe_ctx, @@ -1627,6 +1598,22 @@ static enum dc_status dce110_enable_stream_timing( return DC_OK; } +static void +dce110_select_crtc_source(struct pipe_ctx *pipe_ctx) +{ + struct dc_link *link = pipe_ctx->stream->link; + struct dc_bios *bios = link->ctx->dc_bios; + struct bp_crtc_source_select crtc_source_select = {0}; + enum engine_id engine_id = link->link_enc->preferred_engine; + + if (dc_is_rgb_signal(pipe_ctx->stream->signal)) + engine_id = link->link_enc->analog_engine; + crtc_source_select.controller_id = CONTROLLER_ID_D0 + pipe_ctx->stream_res.tg->inst; + crtc_source_select.color_depth = pipe_ctx->stream->timing.display_color_depth; + crtc_source_select.engine_id = engine_id; + crtc_source_select.sink_signal = pipe_ctx->stream->signal; + bios->funcs->select_crtc_source(bios, &crtc_source_select); +} enum dc_status dce110_apply_single_controller_ctx_to_hw( struct pipe_ctx *pipe_ctx, @@ -1647,6 +1634,10 @@ enum dc_status dce110_apply_single_controller_ctx_to_hw( hws->funcs.disable_stream_gating(dc, pipe_ctx); } + if (pipe_ctx->stream->signal == SIGNAL_TYPE_RGB) { + dce110_select_crtc_source(pipe_ctx); + } + if (pipe_ctx->stream_res.audio != NULL) { struct audio_output audio_output = {0}; @@ -1726,7 +1717,8 @@ enum dc_status dce110_apply_single_controller_ctx_to_hw( pipe_ctx->stream_res.tg->funcs->set_static_screen_control( pipe_ctx->stream_res.tg, event_triggers, 2); - if (!dc_is_virtual_signal(pipe_ctx->stream->signal)) + if (!dc_is_virtual_signal(pipe_ctx->stream->signal) && + !dc_is_rgb_signal(pipe_ctx->stream->signal)) pipe_ctx->stream_res.stream_enc->funcs->dig_connect_to_otg( pipe_ctx->stream_res.stream_enc, pipe_ctx->stream_res.tg->inst); @@ -1948,6 +1940,35 @@ static void clean_up_dsc_blocks(struct dc *dc) } } +static void dc_hwss_enable_otg_pwa( + struct dc *dc, + struct pipe_ctx *pipe_ctx) +{ + struct timing_generator *tg = NULL; + + if (dc->debug.enable_otg_frame_sync_pwa == 0) + return; + + if (pipe_ctx == NULL || pipe_ctx->stream_res.tg == NULL) + return; + tg = pipe_ctx->stream_res.tg; + + /*only enable this if one active*/ + if (tg->funcs->enable_otg_pwa) { + struct otc_pwa_frame_sync pwa_param = {0}; + + DC_LOGGER_INIT(dc->ctx); + /* mode 1 to choose generate pwa sync signal on line 0 counting + * from vstartup at very beginning of the frame + */ + pwa_param.pwa_frame_sync_line_offset = 0; + pwa_param.pwa_sync_mode = DC_OTG_PWA_FRAME_SYNC_MODE_VSTARTUP; + /*frame sync line for generating high frame sync*/ + tg->funcs->enable_otg_pwa(tg, &pwa_param); + DC_LOG_DC("Enable OTG PWA frame sync on TG %d\n", tg->inst); + } +} + /* * When ASIC goes from VBIOS/VGA mode to driver/accelerated mode we need: * 1. Power down all DC HW blocks @@ -2024,6 +2045,7 @@ void dce110_enable_accelerated_mode(struct dc *dc, struct dc_state *context) // If VBios supports it, we check it from reigster or other flags. pipe_ctx->stream_res.pix_clk_params.dio_se_pix_per_cycle = 1; } + dc_hwss_enable_otg_pwa(dc, pipe_ctx); } break; } @@ -2593,6 +2615,18 @@ enum dc_status dce110_apply_ctx_to_hw( #endif } + + if (dc->debug.enable_otg_frame_sync_pwa && context->stream_count == 1) { + /* only enable this on one OTG*/ + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i]; + + if (pipe_ctx && pipe_ctx->stream != NULL) { + dc_hwss_enable_otg_pwa(dc, pipe_ctx); + break; + } + } + } if (dc->fbc_compressor) enable_fbc(dc, dc->current_state); @@ -3320,15 +3354,6 @@ void dce110_enable_tmds_link_output(struct dc_link *link, link->phy_state.symclk_state = SYMCLK_ON_TX_ON; } -static void dce110_enable_analog_link_output( - struct dc_link *link, - uint32_t pix_clk_100hz) -{ - link->link_enc->funcs->enable_analog_output( - link->link_enc, - pix_clk_100hz); -} - void dce110_enable_dp_link_output( struct dc_link *link, const struct link_resource *link_res, @@ -3376,11 +3401,6 @@ void dce110_enable_dp_link_output( } } - if (link->ext_enc_id.id) { - dce110_external_encoder_control(EXTERNAL_ENCODER_CONTROL_INIT, link, NULL); - dce110_external_encoder_control(EXTERNAL_ENCODER_CONTROL_SETUP, link, NULL); - } - if (dc->link_srv->dp_get_encoding_format(link_settings) == DP_8b_10b_ENCODING) { if (dc->clk_mgr->funcs->notify_link_rate_change) dc->clk_mgr->funcs->notify_link_rate_change(dc->clk_mgr, link); @@ -3471,10 +3491,8 @@ static const struct hw_sequencer_funcs dce110_funcs = { .enable_lvds_link_output = dce110_enable_lvds_link_output, .enable_tmds_link_output = dce110_enable_tmds_link_output, .enable_dp_link_output = dce110_enable_dp_link_output, - .enable_analog_link_output = dce110_enable_analog_link_output, .disable_link_output = dce110_disable_link_output, .dac_load_detect = dce110_dac_load_detect, - .prepare_ddc = dce110_prepare_ddc, }; static const struct hwseq_private_funcs dce110_private_funcs = { diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c index 55c48c4bc6ce..357899116ecd 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c @@ -95,10 +95,6 @@ void dcn401_program_gamut_remap(struct pipe_ctx *pipe_ctx) unsigned int mpcc_id = pipe_ctx->plane_res.mpcc_inst; struct mpc *mpc = pipe_ctx->stream_res.opp->ctx->dc->res_pool->mpc; - //For now assert if location is not pre-blend - if (pipe_ctx->plane_state) - ASSERT(pipe_ctx->plane_state->mcm_location == MPCC_MOVABLE_CM_LOCATION_BEFORE); - // program MPCC_MCM_FIRST_GAMUT_REMAP memset(&mpc_adjust, 0, sizeof(mpc_adjust)); mpc_adjust.gamut_adjust_type = GRAPHICS_GAMUT_ADJUST_TYPE_BYPASS; @@ -304,6 +300,7 @@ void dcn401_init_hw(struct dc *dc) } } } + for (i = 0; i < res_pool->audio_count; i++) { struct audio *audio = res_pool->audios[i]; @@ -372,293 +369,179 @@ void dcn401_init_hw(struct dc *dc) } } -static void dcn401_get_mcm_lut_xable_from_pipe_ctx(struct dc *dc, struct pipe_ctx *pipe_ctx, - enum MCM_LUT_XABLE *shaper_xable, - enum MCM_LUT_XABLE *lut3d_xable, - enum MCM_LUT_XABLE *lut1d_xable) +void dcn401_trigger_3dlut_dma_load(struct dc *dc, struct pipe_ctx *pipe_ctx) { - enum dc_cm2_shaper_3dlut_setting shaper_3dlut_setting = DC_CM2_SHAPER_3DLUT_SETTING_BYPASS_ALL; - bool lut1d_enable = false; - struct mpc *mpc = dc->res_pool->mpc; - int mpcc_id = pipe_ctx->plane_res.hubp->inst; - - if (!pipe_ctx->plane_state) - return; - shaper_3dlut_setting = pipe_ctx->plane_state->mcm_shaper_3dlut_setting; - lut1d_enable = pipe_ctx->plane_state->mcm_lut1d_enable; - mpc->funcs->set_movable_cm_location(mpc, MPCC_MOVABLE_CM_LOCATION_BEFORE, mpcc_id); - pipe_ctx->plane_state->mcm_location = MPCC_MOVABLE_CM_LOCATION_BEFORE; - - *lut1d_xable = lut1d_enable ? MCM_LUT_ENABLE : MCM_LUT_DISABLE; + struct hubp *hubp = pipe_ctx->plane_res.hubp; - switch (shaper_3dlut_setting) { - case DC_CM2_SHAPER_3DLUT_SETTING_BYPASS_ALL: - *lut3d_xable = *shaper_xable = MCM_LUT_DISABLE; - break; - case DC_CM2_SHAPER_3DLUT_SETTING_ENABLE_SHAPER: - *lut3d_xable = MCM_LUT_DISABLE; - *shaper_xable = MCM_LUT_ENABLE; - break; - case DC_CM2_SHAPER_3DLUT_SETTING_ENABLE_SHAPER_3DLUT: - *lut3d_xable = *shaper_xable = MCM_LUT_ENABLE; - break; + if (hubp->funcs->hubp_enable_3dlut_fl) { + hubp->funcs->hubp_enable_3dlut_fl(hubp, true); } } -void dcn401_populate_mcm_luts(struct dc *dc, - struct pipe_ctx *pipe_ctx, - struct dc_cm2_func_luts mcm_luts, - bool lut_bank_a) +bool dcn401_set_mcm_luts(struct pipe_ctx *pipe_ctx, + const struct dc_plane_state *plane_state) { + struct dc *dc = pipe_ctx->plane_res.hubp->ctx->dc; struct dpp *dpp_base = pipe_ctx->plane_res.dpp; struct hubp *hubp = pipe_ctx->plane_res.hubp; + const struct dc_plane_cm *cm = &plane_state->cm; int mpcc_id = hubp->inst; struct mpc *mpc = dc->res_pool->mpc; union mcm_lut_params m_lut_params; - enum dc_cm2_transfer_func_source lut3d_src = mcm_luts.lut3d_data.lut3d_src; - enum hubp_3dlut_fl_format format = 0; - enum hubp_3dlut_fl_mode mode; - enum hubp_3dlut_fl_width width = 0; - enum hubp_3dlut_fl_addressing_mode addr_mode; - enum hubp_3dlut_fl_crossbar_bit_slice crossbar_bit_slice_y_g = 0; - enum hubp_3dlut_fl_crossbar_bit_slice crossbar_bit_slice_cb_b = 0; - enum hubp_3dlut_fl_crossbar_bit_slice crossbar_bit_slice_cr_r = 0; - enum MCM_LUT_XABLE shaper_xable = MCM_LUT_DISABLE; - enum MCM_LUT_XABLE lut3d_xable = MCM_LUT_DISABLE; - enum MCM_LUT_XABLE lut1d_xable = MCM_LUT_DISABLE; + struct dc_3dlut_dma lut3d_dma; + bool lut_enable; + bool lut_bank_a; bool rval; + bool result = true; - dcn401_get_mcm_lut_xable_from_pipe_ctx(dc, pipe_ctx, &shaper_xable, &lut3d_xable, &lut1d_xable); + /* decide LUT bank based on current in use */ + mpc->funcs->get_lut_mode(mpc, MCM_LUT_1DLUT, mpcc_id, &lut_enable, &lut_bank_a); + if (!lut_enable) { + mpc->funcs->get_lut_mode(mpc, MCM_LUT_SHAPER, mpcc_id, &lut_enable, &lut_bank_a); + } + if (!lut_enable) { + mpc->funcs->get_lut_mode(mpc, MCM_LUT_3DLUT, mpcc_id, &lut_enable, &lut_bank_a); + } + + /* switch to the next bank */ + if (lut_enable) { + lut_bank_a = !lut_bank_a; + } + + /* MCM location fixed to pre-blend */ + mpc->funcs->set_movable_cm_location(mpc, MPCC_MOVABLE_CM_LOCATION_BEFORE, mpcc_id); /* 1D LUT */ - if (mcm_luts.lut1d_func) { - memset(&m_lut_params, 0, sizeof(m_lut_params)); - if (mcm_luts.lut1d_func->type == TF_TYPE_HWPWL) - m_lut_params.pwl = &mcm_luts.lut1d_func->pwl; - else if (mcm_luts.lut1d_func->type == TF_TYPE_DISTRIBUTED_POINTS) { - rval = cm3_helper_translate_curve_to_hw_format(mpc->ctx, - mcm_luts.lut1d_func, - &dpp_base->regamma_params, false); + lut_enable = cm->flags.bits.blend_enable; + memset(&m_lut_params, 0, sizeof(m_lut_params)); + if (lut_enable) { + if (cm->blend_func.type == TF_TYPE_HWPWL) + m_lut_params.pwl = &cm->blend_func.pwl; + else if (cm->blend_func.type == TF_TYPE_DISTRIBUTED_POINTS) { + rval = cm3_helper_translate_curve_to_hw_format(plane_state->ctx, + &cm->blend_func, + &dpp_base->regamma_params, + false); m_lut_params.pwl = rval ? &dpp_base->regamma_params : NULL; } - if (m_lut_params.pwl) { - if (mpc->funcs->populate_lut) - mpc->funcs->populate_lut(mpc, MCM_LUT_1DLUT, m_lut_params, lut_bank_a, mpcc_id); + + if (!m_lut_params.pwl) { + lut_enable = false; } - if (mpc->funcs->program_lut_mode) - mpc->funcs->program_lut_mode(mpc, MCM_LUT_1DLUT, lut1d_xable && m_lut_params.pwl, lut_bank_a, mpcc_id); + } else { + lut_enable = false; } + if (mpc->funcs->program_lut_mode) + mpc->funcs->program_lut_mode(mpc, MCM_LUT_1DLUT, lut_enable, lut_bank_a, CM_LUT_SIZE_NONE, mpcc_id); + if (lut_enable && mpc->funcs->populate_lut) + mpc->funcs->populate_lut(mpc, MCM_LUT_1DLUT, &m_lut_params, lut_bank_a, mpcc_id); + /* Shaper */ - if (mcm_luts.shaper && mcm_luts.lut3d_data.mpc_3dlut_enable) { + lut_enable = cm->flags.bits.shaper_enable; + if (lut_enable) { memset(&m_lut_params, 0, sizeof(m_lut_params)); - if (mcm_luts.shaper->type == TF_TYPE_HWPWL) - m_lut_params.pwl = &mcm_luts.shaper->pwl; - else if (mcm_luts.shaper->type == TF_TYPE_DISTRIBUTED_POINTS) { + if (cm->shaper_func.type == TF_TYPE_HWPWL) + m_lut_params.pwl = &cm->shaper_func.pwl; + else if (cm->shaper_func.type == TF_TYPE_DISTRIBUTED_POINTS) { ASSERT(false); - rval = cm3_helper_translate_curve_to_hw_format(mpc->ctx, - mcm_luts.shaper, - &dpp_base->regamma_params, true); - m_lut_params.pwl = rval ? &dpp_base->regamma_params : NULL; + rval = cm3_helper_translate_curve_to_hw_format(plane_state->ctx, + &cm->shaper_func, + &dpp_base->shaper_params, + true); + m_lut_params.pwl = rval ? &dpp_base->shaper_params : NULL; } - if (m_lut_params.pwl) { - if (mpc->funcs->mcm.populate_lut) - mpc->funcs->mcm.populate_lut(mpc, m_lut_params, lut_bank_a, mpcc_id); - if (mpc->funcs->program_lut_mode) - mpc->funcs->program_lut_mode(mpc, MCM_LUT_SHAPER, MCM_LUT_ENABLE, lut_bank_a, mpcc_id); + if (!m_lut_params.pwl) { + lut_enable = false; } + } else { + lut_enable = false; } - /* 3DLUT */ - switch (lut3d_src) { - case DC_CM2_TRANSFER_FUNC_SOURCE_SYSMEM: - memset(&m_lut_params, 0, sizeof(m_lut_params)); - if (hubp->funcs->hubp_enable_3dlut_fl) - hubp->funcs->hubp_enable_3dlut_fl(hubp, false); - - if (mcm_luts.lut3d_data.lut3d_func && mcm_luts.lut3d_data.lut3d_func->state.bits.initialized) { - m_lut_params.lut3d = &mcm_luts.lut3d_data.lut3d_func->lut_3d; - if (mpc->funcs->populate_lut) - mpc->funcs->populate_lut(mpc, MCM_LUT_3DLUT, m_lut_params, lut_bank_a, mpcc_id); - if (mpc->funcs->program_lut_mode) - mpc->funcs->program_lut_mode(mpc, MCM_LUT_3DLUT, lut3d_xable, lut_bank_a, - mpcc_id); - } - break; - case DC_CM2_TRANSFER_FUNC_SOURCE_VIDMEM: - switch (mcm_luts.lut3d_data.gpu_mem_params.size) { - case DC_CM2_GPU_MEM_SIZE_171717: - width = hubp_3dlut_fl_width_17; - break; - case DC_CM2_GPU_MEM_SIZE_TRANSFORMED: - width = hubp_3dlut_fl_width_transformed; - break; - default: - //TODO: handle default case - break; - } + if (mpc->funcs->program_lut_mode) + mpc->funcs->program_lut_mode(mpc, MCM_LUT_SHAPER, lut_enable, lut_bank_a, CM_LUT_SIZE_NONE, mpcc_id); + if (lut_enable && mpc->funcs->populate_lut) + mpc->funcs->populate_lut(mpc, MCM_LUT_SHAPER, &m_lut_params, lut_bank_a, mpcc_id); - //check for support - if (mpc->funcs->mcm.is_config_supported && - !mpc->funcs->mcm.is_config_supported(width)) - break; + /* NOTE: Toggling from DMA->Host is not supported atomically as hardware + * blocks writes until 3DLUT FL mode is cleared from HUBP on VUpdate. + * Expectation is either option is used consistently. + */ - if (mpc->funcs->program_lut_read_write_control) - mpc->funcs->program_lut_read_write_control(mpc, MCM_LUT_3DLUT, lut_bank_a, mpcc_id); + /* 3DLUT */ + lut_enable = cm->flags.bits.lut3d_enable; + if (lut_enable && cm->flags.bits.lut3d_dma_enable) { + /* Fast (DMA) Load Mode */ + /* MPC */ if (mpc->funcs->program_lut_mode) - mpc->funcs->program_lut_mode(mpc, MCM_LUT_3DLUT, lut3d_xable, lut_bank_a, mpcc_id); - - if (hubp->funcs->hubp_program_3dlut_fl_addr) - hubp->funcs->hubp_program_3dlut_fl_addr(hubp, mcm_luts.lut3d_data.gpu_mem_params.addr); - - if (mpc->funcs->mcm.program_bit_depth) - mpc->funcs->mcm.program_bit_depth(mpc, mcm_luts.lut3d_data.gpu_mem_params.bit_depth, mpcc_id); + mpc->funcs->program_lut_mode(mpc, MCM_LUT_3DLUT, lut_enable, lut_bank_a, cm->lut3d_dma.size, mpcc_id); - switch (mcm_luts.lut3d_data.gpu_mem_params.layout) { - case DC_CM2_GPU_MEM_LAYOUT_3D_SWIZZLE_LINEAR_RGB: - mode = hubp_3dlut_fl_mode_native_1; - addr_mode = hubp_3dlut_fl_addressing_mode_sw_linear; - break; - case DC_CM2_GPU_MEM_LAYOUT_3D_SWIZZLE_LINEAR_BGR: - mode = hubp_3dlut_fl_mode_native_2; - addr_mode = hubp_3dlut_fl_addressing_mode_sw_linear; - break; - case DC_CM2_GPU_MEM_LAYOUT_1D_PACKED_LINEAR: - mode = hubp_3dlut_fl_mode_transform; - addr_mode = hubp_3dlut_fl_addressing_mode_simple_linear; - break; - default: - mode = hubp_3dlut_fl_mode_disable; - addr_mode = hubp_3dlut_fl_addressing_mode_sw_linear; - break; - } - if (hubp->funcs->hubp_program_3dlut_fl_mode) - hubp->funcs->hubp_program_3dlut_fl_mode(hubp, mode); + /* only supports 12 bit */ + if (mpc->funcs->program_lut_read_write_control) + mpc->funcs->program_lut_read_write_control(mpc, MCM_LUT_3DLUT, lut_bank_a, 12, mpcc_id); - if (hubp->funcs->hubp_program_3dlut_fl_addressing_mode) - hubp->funcs->hubp_program_3dlut_fl_addressing_mode(hubp, addr_mode); + if (mpc->funcs->update_3dlut_fast_load_select) + mpc->funcs->update_3dlut_fast_load_select(mpc, mpcc_id, hubp->inst); - switch (mcm_luts.lut3d_data.gpu_mem_params.format_params.format) { - case DC_CM2_GPU_MEM_FORMAT_16161616_UNORM_12MSB: - format = hubp_3dlut_fl_format_unorm_12msb_bitslice; - break; - case DC_CM2_GPU_MEM_FORMAT_16161616_UNORM_12LSB: - format = hubp_3dlut_fl_format_unorm_12lsb_bitslice; - break; - case DC_CM2_GPU_MEM_FORMAT_16161616_FLOAT_FP1_5_10: - format = hubp_3dlut_fl_format_float_fp1_5_10; - break; - } - if (hubp->funcs->hubp_program_3dlut_fl_format) - hubp->funcs->hubp_program_3dlut_fl_format(hubp, format); - if (hubp->funcs->hubp_update_3dlut_fl_bias_scale && - mpc->funcs->mcm.program_bias_scale) { - mpc->funcs->mcm.program_bias_scale(mpc, - mcm_luts.lut3d_data.gpu_mem_params.format_params.float_params.bias, - mcm_luts.lut3d_data.gpu_mem_params.format_params.float_params.scale, - mpcc_id); - hubp->funcs->hubp_update_3dlut_fl_bias_scale(hubp, - mcm_luts.lut3d_data.gpu_mem_params.format_params.float_params.bias, - mcm_luts.lut3d_data.gpu_mem_params.format_params.float_params.scale); - } - - //navi 4x has a bug and r and blue are swapped and need to be worked around here in - //TODO: need to make a method for get_xbar per asic OR do the workaround in program_crossbar for 4x - switch (mcm_luts.lut3d_data.gpu_mem_params.component_order) { - case DC_CM2_GPU_MEM_PIXEL_COMPONENT_ORDER_RGBA: - default: - crossbar_bit_slice_cr_r = hubp_3dlut_fl_crossbar_bit_slice_0_15; - crossbar_bit_slice_y_g = hubp_3dlut_fl_crossbar_bit_slice_16_31; - crossbar_bit_slice_cb_b = hubp_3dlut_fl_crossbar_bit_slice_32_47; - break; - } + /* HUBP */ + if (hubp->funcs->hubp_program_3dlut_fl_config) + hubp->funcs->hubp_program_3dlut_fl_config(hubp, &cm->lut3d_dma); if (hubp->funcs->hubp_program_3dlut_fl_crossbar) - hubp->funcs->hubp_program_3dlut_fl_crossbar(hubp, - crossbar_bit_slice_cr_r, - crossbar_bit_slice_y_g, - crossbar_bit_slice_cb_b); - - if (mpc->funcs->mcm.program_lut_read_write_control) - mpc->funcs->mcm.program_lut_read_write_control(mpc, MCM_LUT_3DLUT, lut_bank_a, true, mpcc_id); + hubp->funcs->hubp_program_3dlut_fl_crossbar(hubp, cm->lut3d_dma.format); - if (mpc->funcs->mcm.program_3dlut_size) - mpc->funcs->mcm.program_3dlut_size(mpc, width, mpcc_id); - - if (mpc->funcs->update_3dlut_fast_load_select) - mpc->funcs->update_3dlut_fast_load_select(mpc, mpcc_id, hubp->inst); + if (hubp->funcs->hubp_program_3dlut_fl_addr) + hubp->funcs->hubp_program_3dlut_fl_addr(hubp, &cm->lut3d_dma.addr); - if (hubp->funcs->hubp_enable_3dlut_fl) + if (hubp->funcs->hubp_enable_3dlut_fl) { hubp->funcs->hubp_enable_3dlut_fl(hubp, true); - else { - if (mpc->funcs->program_lut_mode) { - mpc->funcs->program_lut_mode(mpc, MCM_LUT_SHAPER, MCM_LUT_DISABLE, lut_bank_a, mpcc_id); - mpc->funcs->program_lut_mode(mpc, MCM_LUT_3DLUT, MCM_LUT_DISABLE, lut_bank_a, mpcc_id); - mpc->funcs->program_lut_mode(mpc, MCM_LUT_1DLUT, MCM_LUT_DISABLE, lut_bank_a, mpcc_id); - } + } else { + /* GPU memory only supports fast load path */ + BREAK_TO_DEBUGGER(); + lut_enable = false; + result = false; } - break; + } else { + /* Legacy (Host) Load Mode */ + memset(&m_lut_params, 0, sizeof(m_lut_params)); - } -} + if (cm->flags.bits.lut3d_enable && cm->lut3d_func.state.bits.initialized) { + m_lut_params.lut3d = &cm->lut3d_func.lut_3d; + } else { + lut_enable = false; + } -void dcn401_trigger_3dlut_dma_load(struct dc *dc, struct pipe_ctx *pipe_ctx) -{ - struct hubp *hubp = pipe_ctx->plane_res.hubp; + /* MPC */ + if (mpc->funcs->program_lut_mode) + mpc->funcs->program_lut_mode(mpc, + MCM_LUT_3DLUT, + lut_enable, + lut_bank_a, + cm->lut3d_func.lut_3d.use_tetrahedral_9 ? CM_LUT_SIZE_999 : CM_LUT_SIZE_171717, + mpcc_id); + + if (lut_enable) { + if (mpc->funcs->program_lut_read_write_control) + mpc->funcs->program_lut_read_write_control(mpc, + MCM_LUT_3DLUT, + lut_bank_a, + cm->lut3d_func.lut_3d.use_12bits ? 12 : 10, + mpcc_id); - if (hubp->funcs->hubp_enable_3dlut_fl) { - hubp->funcs->hubp_enable_3dlut_fl(hubp, true); - } -} + if (mpc->funcs->update_3dlut_fast_load_select) + mpc->funcs->update_3dlut_fast_load_select(mpc, mpcc_id, 0xf); -bool dcn401_set_mcm_luts(struct pipe_ctx *pipe_ctx, - const struct dc_plane_state *plane_state) -{ - struct dpp *dpp_base = pipe_ctx->plane_res.dpp; - int mpcc_id = pipe_ctx->plane_res.hubp->inst; - struct dc *dc = pipe_ctx->stream_res.opp->ctx->dc; - struct mpc *mpc = dc->res_pool->mpc; - bool result; - const struct pwl_params *lut_params = NULL; - bool rval; + if (mpc->funcs->populate_lut) + mpc->funcs->populate_lut(mpc, MCM_LUT_3DLUT, &m_lut_params, lut_bank_a, mpcc_id); + } - if (plane_state->mcm_luts.lut3d_data.lut3d_src == DC_CM2_TRANSFER_FUNC_SOURCE_VIDMEM) { - dcn401_populate_mcm_luts(dc, pipe_ctx, plane_state->mcm_luts, plane_state->lut_bank_a); - return true; - } + /* HUBP */ + memset(&lut3d_dma, 0, sizeof(lut3d_dma)); + if (hubp->funcs->hubp_program_3dlut_fl_config) + hubp->funcs->hubp_program_3dlut_fl_config(hubp, &lut3d_dma); - mpc->funcs->set_movable_cm_location(mpc, MPCC_MOVABLE_CM_LOCATION_BEFORE, mpcc_id); - pipe_ctx->plane_state->mcm_location = MPCC_MOVABLE_CM_LOCATION_BEFORE; - // 1D LUT - if (plane_state->blend_tf.type == TF_TYPE_HWPWL) - lut_params = &plane_state->blend_tf.pwl; - else if (plane_state->blend_tf.type == TF_TYPE_DISTRIBUTED_POINTS) { - rval = cm3_helper_translate_curve_to_hw_format(plane_state->ctx, - &plane_state->blend_tf, - &dpp_base->regamma_params, false); - lut_params = rval ? &dpp_base->regamma_params : NULL; - } - result = mpc->funcs->program_1dlut(mpc, lut_params, mpcc_id); - lut_params = NULL; - - // Shaper - if (plane_state->in_shaper_func.type == TF_TYPE_HWPWL) - lut_params = &plane_state->in_shaper_func.pwl; - else if (plane_state->in_shaper_func.type == TF_TYPE_DISTRIBUTED_POINTS) { - // TODO: dpp_base replace - rval = cm3_helper_translate_curve_to_hw_format(plane_state->ctx, - &plane_state->in_shaper_func, - &dpp_base->shaper_params, true); - lut_params = rval ? &dpp_base->shaper_params : NULL; - } - result &= mpc->funcs->program_shaper(mpc, lut_params, mpcc_id); - - // 3D - if (mpc->funcs->program_3dlut) { - if (plane_state->lut3d_func.state.bits.initialized == 1) - result &= mpc->funcs->program_3dlut(mpc, &plane_state->lut3d_func.lut_3d, mpcc_id); - else - result &= mpc->funcs->program_3dlut(mpc, NULL, mpcc_id); + if (hubp->funcs->hubp_enable_3dlut_fl) + hubp->funcs->hubp_enable_3dlut_fl(hubp, false); } return result; @@ -982,6 +865,8 @@ void dcn401_enable_stream(struct pipe_ctx *pipe_ctx) } } + link_hwss->setup_stream_attribute(pipe_ctx); + if (dc->res_pool->dccg->funcs->set_pixel_rate_div) { dc->res_pool->dccg->funcs->set_pixel_rate_div( dc->res_pool->dccg, @@ -1930,10 +1815,9 @@ void dcn401_perform_3dlut_wa_unlock(struct pipe_ctx *pipe_ctx) for (odm_pipe = pipe_ctx; odm_pipe != NULL; odm_pipe = odm_pipe->next_odm_pipe) { for (mpc_pipe = odm_pipe; mpc_pipe != NULL; mpc_pipe = mpc_pipe->bottom_pipe) { - if (mpc_pipe->plane_state && mpc_pipe->plane_state->mcm_luts.lut3d_data.lut3d_src - == DC_CM2_TRANSFER_FUNC_SOURCE_VIDMEM - && mpc_pipe->plane_state->mcm_shaper_3dlut_setting - == DC_CM2_SHAPER_3DLUT_SETTING_ENABLE_SHAPER_3DLUT) { + if (mpc_pipe->plane_state && + mpc_pipe->plane_state->cm.flags.bits.lut3d_enable && + mpc_pipe->plane_state->cm.flags.bits.lut3d_dma_enable) { wa_pipes[wa_pipe_ct++] = mpc_pipe; } } @@ -2060,6 +1944,22 @@ void dcn401_reset_back_end_for_pipe( DC_LOG_DEBUG("Reset back end for pipe %d, tg:%d\n", pipe_ctx->pipe_idx, pipe_ctx->stream_res.tg->inst); } +static void dc_hwss_disable_otg_pwa(struct dc *dc) +{ + if (dc->debug.enable_otg_frame_sync_pwa) { + int i; + + /*reset all the otg*/ + for (i = dc->res_pool->timing_generator_count - 1; i >= 0 ; i--) { + struct timing_generator *tg = dc->res_pool->timing_generators[i]; + + if (tg->funcs->disable_otg_pwa) { + tg->funcs->disable_otg_pwa(tg); + DC_LOG_DC("otg frame sync pwa disabled on otg%d\n", tg->inst); + } + } + } +} void dcn401_reset_hw_ctx_wrap( struct dc *dc, @@ -2068,6 +1968,7 @@ void dcn401_reset_hw_ctx_wrap( int i; struct dce_hwseq *hws = dc->hwseq; + dc_hwss_disable_otg_pwa(dc); /* Reset Back End*/ for (i = dc->res_pool->pipe_count - 1; i >= 0 ; i--) { struct pipe_ctx *pipe_ctx_old = @@ -2258,6 +2159,10 @@ void dcn401_program_pipe( pipe_ctx->stream_res.test_pattern_params.height, pipe_ctx->stream_res.test_pattern_params.offset); } + if (pipe_ctx->plane_state + && pipe_ctx->plane_state->update_flags.bits.cm_hist_change + && hws->funcs.program_cm_hist) + hws->funcs.program_cm_hist(dc, pipe_ctx, pipe_ctx->plane_state); } /* @@ -2409,6 +2314,13 @@ void dcn401_program_pipe_sequence( pipe_ctx->stream_res.test_pattern_params.offset); } + if (pipe_ctx->plane_state + && pipe_ctx->plane_state->update_flags.bits.cm_hist_change + && hws->funcs.program_cm_hist) { + + hwss_add_dpp_program_cm_hist(seq_state, pipe_ctx->plane_res.dpp, + pipe_ctx->plane_state->cm_hist_control, pipe_ctx->plane_state->color_space); + } } void dcn401_program_front_end_for_ctx( diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.c index f0e1ed0f2949..8e12dc1297c4 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.c @@ -261,7 +261,7 @@ void dcn42_init_hw(struct dc *dc) dc->res_pool->hubbub->funcs->init_crb(dc->res_pool->hubbub); if (dc->res_pool->hubbub->funcs->set_request_limit && dc->config.sdpif_request_limit_words_per_umc > 0) - dc->res_pool->hubbub->funcs->set_request_limit(dc->res_pool->hubbub, dc->ctx->dc_bios->vram_info.num_chans, dc->config.sdpif_request_limit_words_per_umc); + dc->res_pool->hubbub->funcs->set_request_limit(dc->res_pool->hubbub, dc->clk_mgr->bw_params->num_channels, dc->config.sdpif_request_limit_words_per_umc); // Get DMCUB capabilities if (dc->ctx->dmub_srv) { @@ -275,8 +275,7 @@ void dcn42_init_hw(struct dc *dc) || res_pool->ref_clocks.dchub_ref_clock_inKhz / 1000 != current_dchub_ref_freq) { /* update bounding box if FAMS2 disabled, or if dchub clk has changed */ if (dc->clk_mgr) - dc->res_pool->funcs->update_bw_bounding_box(dc, - dc->clk_mgr->bw_params); + dc->res_pool->funcs->update_bw_bounding_box(dc, dc->clk_mgr->bw_params); } } if (dc->res_pool->pg_cntl) { @@ -383,68 +382,6 @@ void dcn42_program_cm_hist( plane_state->cm_hist_control, plane_state->color_space); } -static void dc_get_lut_xbar( - enum dc_cm2_gpu_mem_pixel_component_order order, - enum hubp_3dlut_fl_crossbar_bit_slice *cr_r, - enum hubp_3dlut_fl_crossbar_bit_slice *y_g, - enum hubp_3dlut_fl_crossbar_bit_slice *cb_b) -{ - switch (order) { - case DC_CM2_GPU_MEM_PIXEL_COMPONENT_ORDER_RGBA: - *cr_r = hubp_3dlut_fl_crossbar_bit_slice_32_47; - *y_g = hubp_3dlut_fl_crossbar_bit_slice_16_31; - *cb_b = hubp_3dlut_fl_crossbar_bit_slice_0_15; - break; - case DC_CM2_GPU_MEM_PIXEL_COMPONENT_ORDER_BGRA: - *cr_r = hubp_3dlut_fl_crossbar_bit_slice_0_15; - *y_g = hubp_3dlut_fl_crossbar_bit_slice_16_31; - *cb_b = hubp_3dlut_fl_crossbar_bit_slice_32_47; - break; - } -} - -static void dc_get_lut_mode( - enum dc_cm2_gpu_mem_layout layout, - enum hubp_3dlut_fl_mode *mode, - enum hubp_3dlut_fl_addressing_mode *addr_mode) -{ - switch (layout) { - case DC_CM2_GPU_MEM_LAYOUT_3D_SWIZZLE_LINEAR_RGB: - *mode = hubp_3dlut_fl_mode_native_1; - *addr_mode = hubp_3dlut_fl_addressing_mode_sw_linear; - break; - case DC_CM2_GPU_MEM_LAYOUT_3D_SWIZZLE_LINEAR_BGR: - *mode = hubp_3dlut_fl_mode_native_2; - *addr_mode = hubp_3dlut_fl_addressing_mode_sw_linear; - break; - case DC_CM2_GPU_MEM_LAYOUT_1D_PACKED_LINEAR: - *mode = hubp_3dlut_fl_mode_transform; - *addr_mode = hubp_3dlut_fl_addressing_mode_simple_linear; - break; - default: - *mode = hubp_3dlut_fl_mode_disable; - *addr_mode = hubp_3dlut_fl_addressing_mode_sw_linear; - break; - } -} - -static void dc_get_lut_format( - enum dc_cm2_gpu_mem_format dc_format, - enum hubp_3dlut_fl_format *format) -{ - switch (dc_format) { - case DC_CM2_GPU_MEM_FORMAT_16161616_UNORM_12MSB: - *format = hubp_3dlut_fl_format_unorm_12msb_bitslice; - break; - case DC_CM2_GPU_MEM_FORMAT_16161616_UNORM_12LSB: - *format = hubp_3dlut_fl_format_unorm_12lsb_bitslice; - break; - case DC_CM2_GPU_MEM_FORMAT_16161616_FLOAT_FP1_5_10: - *format = hubp_3dlut_fl_format_float_fp1_5_10; - break; - } -} - static bool dc_is_rmcm_3dlut_supported(struct hubp *hubp, struct mpc *mpc) { if (mpc->funcs->rmcm.power_on_shaper_3dlut && @@ -455,119 +392,17 @@ static bool dc_is_rmcm_3dlut_supported(struct hubp *hubp, struct mpc *mpc) return false; } -static bool is_rmcm_3dlut_fl_supported(struct dc *dc, enum dc_cm2_gpu_mem_size size) -{ - if (!dc->caps.color.mpc.rmcm_3d_lut_caps.dma_3d_lut) - return false; - if (size == DC_CM2_GPU_MEM_SIZE_171717) - return (dc->caps.color.mpc.rmcm_3d_lut_caps.lut_dim_caps.dim_17); - else if (size == DC_CM2_GPU_MEM_SIZE_333333) - return (dc->caps.color.mpc.rmcm_3d_lut_caps.lut_dim_caps.dim_33); - return false; -} - -static void dcn42_set_mcm_location_post_blend(struct dc *dc, struct pipe_ctx *pipe_ctx, bool bPostBlend) -{ - struct mpc *mpc = dc->res_pool->mpc; - int mpcc_id = pipe_ctx->plane_res.hubp->inst; - - if (!pipe_ctx->plane_state) - return; - - mpc->funcs->set_movable_cm_location(mpc, MPCC_MOVABLE_CM_LOCATION_BEFORE, mpcc_id); - pipe_ctx->plane_state->mcm_location = (bPostBlend) ? - MPCC_MOVABLE_CM_LOCATION_AFTER : - MPCC_MOVABLE_CM_LOCATION_BEFORE; -} - -static void dcn42_get_mcm_lut_xable_from_pipe_ctx(struct dc *dc, struct pipe_ctx *pipe_ctx, - enum MCM_LUT_XABLE *shaper_xable, - enum MCM_LUT_XABLE *lut3d_xable, - enum MCM_LUT_XABLE *lut1d_xable) -{ - enum dc_cm2_shaper_3dlut_setting shaper_3dlut_setting = DC_CM2_SHAPER_3DLUT_SETTING_BYPASS_ALL; - bool lut1d_enable = false; - struct mpc *mpc = dc->res_pool->mpc; - int mpcc_id = pipe_ctx->plane_res.hubp->inst; - - if (!pipe_ctx->plane_state) - return; - shaper_3dlut_setting = pipe_ctx->plane_state->mcm_shaper_3dlut_setting; - lut1d_enable = pipe_ctx->plane_state->mcm_lut1d_enable; - mpc->funcs->set_movable_cm_location(mpc, MPCC_MOVABLE_CM_LOCATION_BEFORE, mpcc_id); - pipe_ctx->plane_state->mcm_location = MPCC_MOVABLE_CM_LOCATION_BEFORE; - - *lut1d_xable = lut1d_enable ? MCM_LUT_ENABLE : MCM_LUT_DISABLE; - - switch (shaper_3dlut_setting) { - case DC_CM2_SHAPER_3DLUT_SETTING_BYPASS_ALL: - *lut3d_xable = *shaper_xable = MCM_LUT_DISABLE; - break; - case DC_CM2_SHAPER_3DLUT_SETTING_ENABLE_SHAPER: - *lut3d_xable = MCM_LUT_DISABLE; - *shaper_xable = MCM_LUT_ENABLE; - break; - case DC_CM2_SHAPER_3DLUT_SETTING_ENABLE_SHAPER_3DLUT: - *lut3d_xable = *shaper_xable = MCM_LUT_ENABLE; - break; - } -} - -static void fl_get_lut_mode( - enum dc_cm2_gpu_mem_layout layout, - enum dc_cm2_gpu_mem_size size, - enum hubp_3dlut_fl_mode *mode, - enum hubp_3dlut_fl_addressing_mode *addr_mode, - enum hubp_3dlut_fl_width *width) -{ - *width = hubp_3dlut_fl_width_17; - - if (size == DC_CM2_GPU_MEM_SIZE_333333) - *width = hubp_3dlut_fl_width_33; - - switch (layout) { - case DC_CM2_GPU_MEM_LAYOUT_3D_SWIZZLE_LINEAR_RGB: - *mode = hubp_3dlut_fl_mode_native_1; - *addr_mode = hubp_3dlut_fl_addressing_mode_sw_linear; - break; - case DC_CM2_GPU_MEM_LAYOUT_3D_SWIZZLE_LINEAR_BGR: - *mode = hubp_3dlut_fl_mode_native_2; - *addr_mode = hubp_3dlut_fl_addressing_mode_sw_linear; - break; - case DC_CM2_GPU_MEM_LAYOUT_1D_PACKED_LINEAR: - *mode = hubp_3dlut_fl_mode_transform; - *addr_mode = hubp_3dlut_fl_addressing_mode_simple_linear; - break; - default: - *mode = hubp_3dlut_fl_mode_disable; - *addr_mode = hubp_3dlut_fl_addressing_mode_sw_linear; - break; - } -} - bool dcn42_program_rmcm_luts( struct hubp *hubp, struct pipe_ctx *pipe_ctx, - enum dc_cm2_transfer_func_source lut3d_src, - struct dc_cm2_func_luts *mcm_luts, + const struct dc_plane_cm *cm, struct mpc *mpc, - bool lut_bank_a, int mpcc_id) { struct dpp *dpp_base = pipe_ctx->plane_res.dpp; union mcm_lut_params m_lut_params = {0}; - enum MCM_LUT_XABLE shaper_xable, lut3d_xable = MCM_LUT_DISABLE, lut1d_xable; - enum hubp_3dlut_fl_mode mode; - enum hubp_3dlut_fl_addressing_mode addr_mode; - enum hubp_3dlut_fl_format format = hubp_3dlut_fl_format_unorm_12msb_bitslice; - enum hubp_3dlut_fl_crossbar_bit_slice crossbar_bit_slice_y_g = hubp_3dlut_fl_crossbar_bit_slice_16_31; - enum hubp_3dlut_fl_crossbar_bit_slice crossbar_bit_slice_cb_b = hubp_3dlut_fl_crossbar_bit_slice_0_15; - enum hubp_3dlut_fl_crossbar_bit_slice crossbar_bit_slice_cr_r = hubp_3dlut_fl_crossbar_bit_slice_32_47; - enum hubp_3dlut_fl_width width = hubp_3dlut_fl_width_17; - struct dc *dc = hubp->ctx->dc; - struct hubp_fl_3dlut_config fl_config; struct mpc_fl_3dlut_config mpc_fl_config; struct dc_stream_state *stream = pipe_ctx->stream; @@ -575,25 +410,23 @@ bool dcn42_program_rmcm_luts( // true->false when it can be allocated at DI time struct dc_rmcm_3dlut *rmcm_3dlut = dc_stream_get_3dlut_for_stream(dc, stream, false); + bool lut_bank_a = true; // TODO get from HW + //check to see current pipe is part of a stream with allocated rmcm 3dlut if (!rmcm_3dlut) return false; - rmcm_3dlut->protection_bits = mcm_luts->lut3d_data.rmcm_tmz; - - dcn42_get_mcm_lut_xable_from_pipe_ctx(dc, pipe_ctx, &shaper_xable, &lut3d_xable, &lut1d_xable); - /* Shaper */ - if (mcm_luts->shaper) { + if (cm->flags.bits.shaper_enable) { memset(&m_lut_params, 0, sizeof(m_lut_params)); - if (mcm_luts->shaper->type == TF_TYPE_HWPWL) { - m_lut_params.pwl = &mcm_luts->shaper->pwl; - } else if (mcm_luts->shaper->type == TF_TYPE_DISTRIBUTED_POINTS) { + if (cm->shaper_func.type == TF_TYPE_HWPWL) { + m_lut_params.pwl = &cm->shaper_func.pwl; + } else if (cm->shaper_func.type == TF_TYPE_DISTRIBUTED_POINTS) { ASSERT(false); cm_helper_translate_curve_to_hw_format( dc->ctx, - mcm_luts->shaper, + &cm->shaper_func, &dpp_base->shaper_params, true); m_lut_params.pwl = &dpp_base->shaper_params; } @@ -609,58 +442,21 @@ bool dcn42_program_rmcm_luts( } /* 3DLUT */ - switch (lut3d_src) { - case DC_CM2_TRANSFER_FUNC_SOURCE_SYSMEM: + if (!cm->flags.bits.lut3d_dma_enable) { memset(&m_lut_params, 0, sizeof(m_lut_params)); // Don't know what to do in this case. - //case DC_CM2_TRANSFER_FUNC_SOURCE_SYSMEM: - break; - case DC_CM2_TRANSFER_FUNC_SOURCE_VIDMEM: - fl_get_lut_mode(mcm_luts->lut3d_data.gpu_mem_params.layout, - mcm_luts->lut3d_data.gpu_mem_params.size, - &mode, - &addr_mode, - &width); - - if (!dc_is_rmcm_3dlut_supported(hubp, mpc) || - !mpc->funcs->rmcm.is_config_supported( - (width == hubp_3dlut_fl_width_17 || - width == hubp_3dlut_fl_width_transformed) ? 17 : 33)) + } else { + if (!dc_is_rmcm_3dlut_supported(hubp, mpc)) return false; - // setting native or transformed mode, - dc_get_lut_mode(mcm_luts->lut3d_data.gpu_mem_params.layout, &mode, &addr_mode); - //seems to be only for the MCM - dc_get_lut_format(mcm_luts->lut3d_data.gpu_mem_params.format_params.format, &format); - - dc_get_lut_xbar( - mcm_luts->lut3d_data.gpu_mem_params.component_order, - &crossbar_bit_slice_cr_r, - &crossbar_bit_slice_y_g, - &crossbar_bit_slice_cb_b); - - fl_config.mode = mode; - fl_config.enabled = lut3d_xable != MCM_LUT_DISABLE; - fl_config.address = mcm_luts->lut3d_data.gpu_mem_params.addr; - fl_config.format = format; - fl_config.crossbar_bit_slice_y_g = crossbar_bit_slice_y_g; - fl_config.crossbar_bit_slice_cb_b = crossbar_bit_slice_cb_b; - fl_config.crossbar_bit_slice_cr_r = crossbar_bit_slice_cr_r; - fl_config.width = width; - fl_config.protection_bits = rmcm_3dlut->protection_bits; - fl_config.addr_mode = addr_mode; - fl_config.layout = mcm_luts->lut3d_data.gpu_mem_params.layout; - fl_config.bias = mcm_luts->lut3d_data.gpu_mem_params.format_params.float_params.bias; - fl_config.scale = mcm_luts->lut3d_data.gpu_mem_params.format_params.float_params.scale; - - mpc_fl_config.enabled = fl_config.enabled; - mpc_fl_config.width = width; + mpc_fl_config.enabled = cm->flags.bits.lut3d_enable; + mpc_fl_config.size = cm->lut3d_dma.size; mpc_fl_config.select_lut_bank_a = lut_bank_a; - mpc_fl_config.bit_depth = mcm_luts->lut3d_data.gpu_mem_params.bit_depth; + mpc_fl_config.bit_depth = 0; mpc_fl_config.hubp_index = hubp->inst; - mpc_fl_config.bias = mcm_luts->lut3d_data.gpu_mem_params.format_params.float_params.bias; - mpc_fl_config.scale = mcm_luts->lut3d_data.gpu_mem_params.format_params.float_params.scale; + mpc_fl_config.bias = cm->lut3d_dma.bias; + mpc_fl_config.scale = cm->lut3d_dma.scale; //1. power down the block mpc->funcs->rmcm.power_on_shaper_3dlut(mpc, mpcc_id, false); @@ -668,268 +464,44 @@ bool dcn42_program_rmcm_luts( //2. program RMCM - 3dlut reg programming mpc->funcs->rmcm.fl_3dlut_configure(mpc, &mpc_fl_config, mpcc_id); - hubp->funcs->hubp_program_3dlut_fl_config(hubp, &fl_config); + /* HUBP */ + if (hubp->funcs->hubp_program_3dlut_fl_config) + hubp->funcs->hubp_program_3dlut_fl_config(hubp, &cm->lut3d_dma); + + if (hubp->funcs->hubp_program_3dlut_fl_addr) + hubp->funcs->hubp_program_3dlut_fl_addr(hubp, &cm->lut3d_dma.addr); //3. power on the block mpc->funcs->rmcm.power_on_shaper_3dlut(mpc, mpcc_id, true); - - break; - default: - return false; } return true; } -void dcn42_populate_mcm_luts(struct dc *dc, - struct pipe_ctx *pipe_ctx, - struct dc_cm2_func_luts mcm_luts, - bool lut_bank_a) +bool dcn42_set_mcm_luts(struct pipe_ctx *pipe_ctx, + const struct dc_plane_state *plane_state) { - struct dpp *dpp_base = pipe_ctx->plane_res.dpp; + struct dc *dc = pipe_ctx->plane_res.hubp->ctx->dc; struct hubp *hubp = pipe_ctx->plane_res.hubp; - int mpcc_id = hubp->inst; + const struct dc_plane_cm *cm = &plane_state->cm; struct mpc *mpc = dc->res_pool->mpc; - union mcm_lut_params m_lut_params; - enum dc_cm2_transfer_func_source lut3d_src = mcm_luts.lut3d_data.lut3d_src; - enum hubp_3dlut_fl_format format = 0; - enum hubp_3dlut_fl_mode mode; - enum hubp_3dlut_fl_width width = 0; - enum hubp_3dlut_fl_addressing_mode addr_mode; - enum hubp_3dlut_fl_crossbar_bit_slice crossbar_bit_slice_y_g = 0; - enum hubp_3dlut_fl_crossbar_bit_slice crossbar_bit_slice_cb_b = 0; - enum hubp_3dlut_fl_crossbar_bit_slice crossbar_bit_slice_cr_r = 0; - enum MCM_LUT_XABLE shaper_xable = MCM_LUT_DISABLE; - enum MCM_LUT_XABLE lut3d_xable = MCM_LUT_DISABLE; - enum MCM_LUT_XABLE lut1d_xable = MCM_LUT_DISABLE; - bool rval; - - dcn42_get_mcm_lut_xable_from_pipe_ctx(dc, pipe_ctx, &shaper_xable, &lut3d_xable, &lut1d_xable); - - //MCM - setting its location (Before/After) blender - //set to post blend (true) - dcn42_set_mcm_location_post_blend( - dc, - pipe_ctx, - mcm_luts.lut3d_data.mpc_mcm_post_blend); - - //RMCM - 3dLUT+Shaper - if (mcm_luts.lut3d_data.rmcm_3dlut_enable && - is_rmcm_3dlut_fl_supported(dc, mcm_luts.lut3d_data.gpu_mem_params.size)) { + int mpcc_id = hubp->inst; + bool result; + + /* MCM */ + result = dcn401_set_mcm_luts(pipe_ctx, plane_state); + + /* RMCM */ + if (cm->flags.bits.rmcm_enable && cm->flags.bits.lut3d_dma_enable) { + /* TODO - move RMCM to its own block */ dcn42_program_rmcm_luts( hubp, pipe_ctx, - lut3d_src, - &mcm_luts, + cm, mpc, - lut_bank_a, mpcc_id); } - /* 1D LUT */ - if (mcm_luts.lut1d_func) { - memset(&m_lut_params, 0, sizeof(m_lut_params)); - if (mcm_luts.lut1d_func->type == TF_TYPE_HWPWL) - m_lut_params.pwl = &mcm_luts.lut1d_func->pwl; - else if (mcm_luts.lut1d_func->type == TF_TYPE_DISTRIBUTED_POINTS) { - rval = cm3_helper_translate_curve_to_hw_format(mpc->ctx, - mcm_luts.lut1d_func, - &dpp_base->regamma_params, false); - m_lut_params.pwl = rval ? &dpp_base->regamma_params : NULL; - } - if (m_lut_params.pwl) { - if (mpc->funcs->populate_lut) - mpc->funcs->populate_lut(mpc, MCM_LUT_1DLUT, m_lut_params, lut_bank_a, mpcc_id); - } - if (mpc->funcs->program_lut_mode) - mpc->funcs->program_lut_mode(mpc, MCM_LUT_1DLUT, lut1d_xable && m_lut_params.pwl, lut_bank_a, mpcc_id); - } - - /* Shaper */ - if (mcm_luts.shaper && mcm_luts.lut3d_data.mpc_3dlut_enable) { - memset(&m_lut_params, 0, sizeof(m_lut_params)); - if (mcm_luts.shaper->type == TF_TYPE_HWPWL) - m_lut_params.pwl = &mcm_luts.shaper->pwl; - else if (mcm_luts.shaper->type == TF_TYPE_DISTRIBUTED_POINTS) { - ASSERT(false); - rval = cm3_helper_translate_curve_to_hw_format(mpc->ctx, - mcm_luts.shaper, - &dpp_base->regamma_params, true); - m_lut_params.pwl = rval ? &dpp_base->regamma_params : NULL; - } - if (m_lut_params.pwl) { - if (mpc->funcs->mcm.populate_lut) - mpc->funcs->mcm.populate_lut(mpc, m_lut_params, lut_bank_a, mpcc_id); - if (mpc->funcs->program_lut_mode) - mpc->funcs->program_lut_mode(mpc, MCM_LUT_SHAPER, MCM_LUT_ENABLE, lut_bank_a, mpcc_id); - } - } - - /* 3DLUT */ - switch (lut3d_src) { - case DC_CM2_TRANSFER_FUNC_SOURCE_SYSMEM: - memset(&m_lut_params, 0, sizeof(m_lut_params)); - if (hubp->funcs->hubp_enable_3dlut_fl) - hubp->funcs->hubp_enable_3dlut_fl(hubp, false); - - if (mcm_luts.lut3d_data.lut3d_func && mcm_luts.lut3d_data.lut3d_func->state.bits.initialized) { - m_lut_params.lut3d = &mcm_luts.lut3d_data.lut3d_func->lut_3d; - if (mpc->funcs->populate_lut) - mpc->funcs->populate_lut(mpc, MCM_LUT_3DLUT, m_lut_params, lut_bank_a, mpcc_id); - if (mpc->funcs->program_lut_mode) - mpc->funcs->program_lut_mode(mpc, MCM_LUT_3DLUT, lut3d_xable, lut_bank_a, - mpcc_id); - } - break; - case DC_CM2_TRANSFER_FUNC_SOURCE_VIDMEM: - switch (mcm_luts.lut3d_data.gpu_mem_params.size) { - case DC_CM2_GPU_MEM_SIZE_333333: - width = hubp_3dlut_fl_width_33; - break; - case DC_CM2_GPU_MEM_SIZE_171717: - width = hubp_3dlut_fl_width_17; - break; - case DC_CM2_GPU_MEM_SIZE_TRANSFORMED: - width = hubp_3dlut_fl_width_transformed; - break; - default: - //TODO: Handle default case - break; - } - - //check for support - if (mpc->funcs->mcm.is_config_supported && - !mpc->funcs->mcm.is_config_supported(width)) - break; - - if (mpc->funcs->program_lut_read_write_control) - mpc->funcs->program_lut_read_write_control(mpc, MCM_LUT_3DLUT, lut_bank_a, mpcc_id); - if (mpc->funcs->program_lut_mode) - mpc->funcs->program_lut_mode(mpc, MCM_LUT_3DLUT, lut3d_xable, lut_bank_a, mpcc_id); - - if (hubp->funcs->hubp_program_3dlut_fl_addr) - hubp->funcs->hubp_program_3dlut_fl_addr(hubp, mcm_luts.lut3d_data.gpu_mem_params.addr); - - if (mpc->funcs->mcm.program_bit_depth) - mpc->funcs->mcm.program_bit_depth(mpc, mcm_luts.lut3d_data.gpu_mem_params.bit_depth, mpcc_id); - - dc_get_lut_mode(mcm_luts.lut3d_data.gpu_mem_params.layout, &mode, &addr_mode); - if (hubp->funcs->hubp_program_3dlut_fl_mode) - hubp->funcs->hubp_program_3dlut_fl_mode(hubp, mode); - - if (hubp->funcs->hubp_program_3dlut_fl_addressing_mode) - hubp->funcs->hubp_program_3dlut_fl_addressing_mode(hubp, addr_mode); - - switch (mcm_luts.lut3d_data.gpu_mem_params.format_params.format) { - case DC_CM2_GPU_MEM_FORMAT_16161616_UNORM_12MSB: - format = hubp_3dlut_fl_format_unorm_12msb_bitslice; - break; - case DC_CM2_GPU_MEM_FORMAT_16161616_UNORM_12LSB: - format = hubp_3dlut_fl_format_unorm_12lsb_bitslice; - break; - case DC_CM2_GPU_MEM_FORMAT_16161616_FLOAT_FP1_5_10: - format = hubp_3dlut_fl_format_float_fp1_5_10; - break; - } - if (hubp->funcs->hubp_program_3dlut_fl_format) - hubp->funcs->hubp_program_3dlut_fl_format(hubp, format); - if (hubp->funcs->hubp_update_3dlut_fl_bias_scale && - mpc->funcs->mcm.program_bias_scale) { - mpc->funcs->mcm.program_bias_scale(mpc, - mcm_luts.lut3d_data.gpu_mem_params.format_params.float_params.bias, - mcm_luts.lut3d_data.gpu_mem_params.format_params.float_params.scale, - mpcc_id); - hubp->funcs->hubp_update_3dlut_fl_bias_scale(hubp, - mcm_luts.lut3d_data.gpu_mem_params.format_params.float_params.bias, - mcm_luts.lut3d_data.gpu_mem_params.format_params.float_params.scale); - } - - //navi 4x has a bug and r and blue are swapped and need to be worked around here in - //TODO: need to make a method for get_xbar per asic OR do the workaround in program_crossbar for 4x - dc_get_lut_xbar( - mcm_luts.lut3d_data.gpu_mem_params.component_order, - &crossbar_bit_slice_cr_r, - &crossbar_bit_slice_y_g, - &crossbar_bit_slice_cb_b); - - if (hubp->funcs->hubp_program_3dlut_fl_crossbar) - hubp->funcs->hubp_program_3dlut_fl_crossbar(hubp, - crossbar_bit_slice_cr_r, - crossbar_bit_slice_y_g, - crossbar_bit_slice_cb_b); - - if (mpc->funcs->mcm.program_lut_read_write_control) - mpc->funcs->mcm.program_lut_read_write_control(mpc, MCM_LUT_3DLUT, lut_bank_a, true, mpcc_id); - - if (mpc->funcs->mcm.program_3dlut_size) - mpc->funcs->mcm.program_3dlut_size(mpc, width, mpcc_id); - - if (mpc->funcs->update_3dlut_fast_load_select) - mpc->funcs->update_3dlut_fast_load_select(mpc, mpcc_id, hubp->inst); - - if (hubp->funcs->hubp_enable_3dlut_fl) - hubp->funcs->hubp_enable_3dlut_fl(hubp, true); - else { - if (mpc->funcs->program_lut_mode) { - mpc->funcs->program_lut_mode(mpc, MCM_LUT_SHAPER, MCM_LUT_DISABLE, lut_bank_a, mpcc_id); - mpc->funcs->program_lut_mode(mpc, MCM_LUT_3DLUT, MCM_LUT_DISABLE, lut_bank_a, mpcc_id); - mpc->funcs->program_lut_mode(mpc, MCM_LUT_1DLUT, MCM_LUT_DISABLE, lut_bank_a, mpcc_id); - } - } - break; - } -} - -bool dcn42_set_mcm_luts(struct pipe_ctx *pipe_ctx, - const struct dc_plane_state *plane_state) -{ - struct dpp *dpp_base = pipe_ctx->plane_res.dpp; - int mpcc_id = pipe_ctx->plane_res.hubp->inst; - struct dc *dc = pipe_ctx->stream_res.opp->ctx->dc; - struct mpc *mpc = dc->res_pool->mpc; - bool result; - const struct pwl_params *lut_params = NULL; - bool rval; - - if (plane_state->mcm_luts.lut3d_data.lut3d_src == DC_CM2_TRANSFER_FUNC_SOURCE_VIDMEM) { - dcn42_populate_mcm_luts(dc, pipe_ctx, plane_state->mcm_luts, plane_state->lut_bank_a); - return true; - } - - mpc->funcs->set_movable_cm_location(mpc, MPCC_MOVABLE_CM_LOCATION_BEFORE, mpcc_id); - pipe_ctx->plane_state->mcm_location = MPCC_MOVABLE_CM_LOCATION_BEFORE; - // 1D LUT - if (plane_state->blend_tf.type == TF_TYPE_HWPWL) - lut_params = &plane_state->blend_tf.pwl; - else if (plane_state->blend_tf.type == TF_TYPE_DISTRIBUTED_POINTS) { - rval = cm3_helper_translate_curve_to_hw_format(plane_state->ctx, - &plane_state->blend_tf, - &dpp_base->regamma_params, false); - lut_params = rval ? &dpp_base->regamma_params : NULL; - } - result = mpc->funcs->program_1dlut(mpc, lut_params, mpcc_id); - lut_params = NULL; - - // Shaper - if (plane_state->in_shaper_func.type == TF_TYPE_HWPWL) - lut_params = &plane_state->in_shaper_func.pwl; - else if (plane_state->in_shaper_func.type == TF_TYPE_DISTRIBUTED_POINTS) { - // TODO: dpp_base replace - rval = cm3_helper_translate_curve_to_hw_format(plane_state->ctx, - &plane_state->in_shaper_func, - &dpp_base->shaper_params, true); - lut_params = rval ? &dpp_base->shaper_params : NULL; - } - result &= mpc->funcs->program_shaper(mpc, lut_params, mpcc_id); - - // 3D - if (mpc->funcs->program_3dlut) { - if (plane_state->lut3d_func.state.bits.initialized == 1) - result &= mpc->funcs->program_3dlut(mpc, &plane_state->lut3d_func.lut_3d, mpcc_id); - else - result &= mpc->funcs->program_3dlut(mpc, NULL, mpcc_id); - } - return result; } void dcn42_hardware_release(struct dc *dc) @@ -1050,7 +622,7 @@ void dcn42_prepare_bandwidth( dc->hwss.hw_block_power_up(dc, &pg_update_state); } - dcn20_prepare_bandwidth(dc, context); + dcn401_prepare_bandwidth(dc, context); } void dcn42_optimize_bandwidth(struct dc *dc, struct dc_state *context) @@ -1470,3 +1042,50 @@ void dcn42_dmub_hw_control_lock_fast(union block_sequence_params *params) dmub_hw_lock_mgr_inbox0_cmd(dc->ctx->dmub_srv, hw_lock_cmd); } } + +/* In headless boot cases, DIG may be turned + * on which causes HW/SW discrepancies. + * To avoid this, power down hardware on boot + * if DIG is turned on + */ +void dcn42_power_down_on_boot(struct dc *dc) +{ + struct dc_link *edp_links[MAX_NUM_EDP]; + struct dc_link *edp_link = NULL; + int edp_num; + int i = 0; + + dc_get_edp_links(dc, edp_links, &edp_num); + if (edp_num) + edp_link = edp_links[0]; + + if (edp_link && edp_link->link_enc->funcs->is_dig_enabled && + edp_link->link_enc->funcs->is_dig_enabled(edp_link->link_enc) && + dc->hwseq->funcs.edp_backlight_control && + dc->hwseq->funcs.power_down && + dc->hwss.edp_power_control) { + dc->hwseq->funcs.edp_backlight_control(edp_link, false); + dc->hwseq->funcs.power_down(dc); + dc->hwss.edp_power_control(edp_link, false); + } else { + for (i = 0; i < dc->link_count; i++) { + struct dc_link *link = dc->links[i]; + + if (link->link_enc && link->link_enc->funcs->is_dig_enabled && + link->link_enc->funcs->is_dig_enabled(link->link_enc) && + dc->hwseq->funcs.power_down) { + dc->hwseq->funcs.power_down(dc); + break; + } + + } + } + + /* + * Call update_clocks with empty context + * to send DISPLAY_OFF + * Otherwise DISPLAY_OFF may not be asserted + */ + if (dc->clk_mgr->funcs->set_low_power_state) + dc->clk_mgr->funcs->set_low_power_state(dc->clk_mgr); +} diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.h b/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.h index 89ebb6520eaf..c4cfeed45b19 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.h +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_hwseq.h @@ -18,18 +18,11 @@ void dcn42_program_cm_hist( bool dcn42_set_mcm_luts(struct pipe_ctx *pipe_ctx, const struct dc_plane_state *plane_state); -void dcn42_populate_mcm_luts(struct dc *dc, - struct pipe_ctx *pipe_ctx, - struct dc_cm2_func_luts mcm_luts, - bool lut_bank_a); - bool dcn42_program_rmcm_luts( struct hubp *hubp, struct pipe_ctx *pipe_ctx, - enum dc_cm2_transfer_func_source lut3d_src, - struct dc_cm2_func_luts *mcm_luts, + const struct dc_plane_cm *cm, struct mpc *mpc, - bool lut_bank_a, int mpcc_id); void dcn42_hardware_release(struct dc *dc); @@ -50,4 +43,5 @@ void dcn42_root_clock_control(struct dc *dc, void dcn42_dmub_hw_control_lock(struct dc *dc, struct dc_state *context, bool lock); void dcn42_dmub_hw_control_lock_fast(union block_sequence_params *params); void dcn42_setup_stereo(struct pipe_ctx *pipe_ctx, struct dc *dc); +void dcn42_power_down_on_boot(struct dc *dc); #endif diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_init.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_init.c index a8e2f59d5e50..b324a2195e8a 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_init.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn42/dcn42_init.c @@ -19,7 +19,7 @@ static const struct hw_sequencer_funcs dcn42_funcs = { .program_gamut_remap = dcn401_program_gamut_remap, .init_hw = dcn42_init_hw, .apply_ctx_to_hw = dce110_apply_ctx_to_hw, - .power_down_on_boot = dcn35_power_down_on_boot, + .power_down_on_boot = dcn42_power_down_on_boot, .apply_ctx_for_surface = NULL, .program_front_end_for_ctx = dcn401_program_front_end_for_ctx, .clear_surface_dcc_and_tiling = dcn10_reset_surface_dcc_and_tiling, @@ -64,6 +64,12 @@ static const struct hw_sequencer_funcs dcn42_funcs = { .set_cursor_position = dcn401_set_cursor_position, .set_cursor_attribute = dcn10_set_cursor_attribute, .set_cursor_sdr_white_level = dcn10_set_cursor_sdr_white_level, + .abort_cursor_offload_update = dcn35_abort_cursor_offload_update, + .begin_cursor_offload_update = dcn35_begin_cursor_offload_update, + .commit_cursor_offload_update = dcn35_commit_cursor_offload_update, + .update_cursor_offload_pipe = dcn401_update_cursor_offload_pipe, + .notify_cursor_offload_drr_update = dcn35_notify_cursor_offload_drr_update, + .program_cursor_offload_now = dcn35_program_cursor_offload_now, .setup_periodic_interrupt = dcn10_setup_periodic_interrupt, .set_clock = dcn10_set_clock, .get_clock = dcn10_get_clock, diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/hubp.h b/drivers/gpu/drm/amd/display/dc/inc/hw/hubp.h index a79019365af8..2a5a81d15950 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/hubp.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/hubp.h @@ -89,7 +89,7 @@ enum hubp_3dlut_fl_addressing_mode { enum hubp_3dlut_fl_width { hubp_3dlut_fl_width_17 = 17, hubp_3dlut_fl_width_33 = 33, - hubp_3dlut_fl_width_transformed = 4916, //mpc default + hubp_3dlut_fl_width_17_transformed = 4916, //mpc default }; enum hubp_3dlut_fl_crossbar_bit_slice { @@ -99,22 +99,6 @@ enum hubp_3dlut_fl_crossbar_bit_slice { hubp_3dlut_fl_crossbar_bit_slice_48_63 = 3 }; -struct hubp_fl_3dlut_config { - bool enabled; - enum hubp_3dlut_fl_width width; - enum hubp_3dlut_fl_mode mode; - enum hubp_3dlut_fl_format format; - uint16_t bias; - uint16_t scale; - struct dc_plane_address address; - enum hubp_3dlut_fl_addressing_mode addr_mode; - enum dc_cm2_gpu_mem_layout layout; - uint8_t protection_bits; - enum hubp_3dlut_fl_crossbar_bit_slice crossbar_bit_slice_y_g; - enum hubp_3dlut_fl_crossbar_bit_slice crossbar_bit_slice_cb_b; - enum hubp_3dlut_fl_crossbar_bit_slice crossbar_bit_slice_cr_r; -}; - struct hubp { const struct hubp_funcs *funcs; struct dc_context *ctx; @@ -243,7 +227,6 @@ struct hubp_funcs { void (*hubp_disable_control)(struct hubp *hubp, bool disable_hubp); unsigned int (*hubp_get_underflow_status)(struct hubp *hubp); void (*hubp_init)(struct hubp *hubp); - void (*dmdata_set_attributes)( struct hubp *hubp, const struct dc_dmdata_attributes *attr); @@ -290,24 +273,15 @@ struct hubp_funcs { void (*hubp_wait_pipe_read_start)(struct hubp *hubp); void (*hubp_program_mcache_id_and_split_coordinate)(struct hubp *hubp, struct dml2_hubp_pipe_mcache_regs *mcache_regs); - void (*hubp_update_3dlut_fl_bias_scale)(struct hubp *hubp, uint16_t bias, uint16_t scale); - void (*hubp_program_3dlut_fl_mode)(struct hubp *hubp, - enum hubp_3dlut_fl_mode mode); - void (*hubp_program_3dlut_fl_format)(struct hubp *hubp, - enum hubp_3dlut_fl_format format); void (*hubp_program_3dlut_fl_addr)(struct hubp *hubp, - const struct dc_plane_address address); + const struct dc_plane_address *address); + void (*hubp_program_3dlut_fl_config)(struct hubp *hubp, + const struct dc_3dlut_dma *config); void (*hubp_program_3dlut_fl_dlg_param)(struct hubp *hubp, int refcyc_per_3dlut_group); void (*hubp_enable_3dlut_fl)(struct hubp *hubp, bool enable); - void (*hubp_program_3dlut_fl_addressing_mode)(struct hubp *hubp, enum hubp_3dlut_fl_addressing_mode addr_mode); - void (*hubp_program_3dlut_fl_width)(struct hubp *hubp, enum hubp_3dlut_fl_width width); - void (*hubp_program_3dlut_fl_tmz_protected)(struct hubp *hubp, uint8_t protection_bits); void (*hubp_program_3dlut_fl_crossbar)(struct hubp *hubp, - enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_y_g, - enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_cb_b, - enum hubp_3dlut_fl_crossbar_bit_slice bit_slice_cr_r); + enum dc_cm_lut_pixel_format format); int (*hubp_get_3dlut_fl_done)(struct hubp *hubp); - void (*hubp_program_3dlut_fl_config)(struct hubp *hubp, struct hubp_fl_3dlut_config *cfg); void (*hubp_clear_tiling)(struct hubp *hubp); uint32_t (*hubp_get_current_read_line)(struct hubp *hubp); uint32_t (*hubp_get_det_config_error)(struct hubp *hubp); diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h b/drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h index a61d12ec61bc..b152f6879495 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h @@ -152,6 +152,13 @@ struct dc_rgb { uint32_t blue; }; +struct tetrahedral_33x33x33 { + struct dc_rgb lut0[8985]; + struct dc_rgb lut1[8984]; + struct dc_rgb lut2[8984]; + struct dc_rgb lut3[8984]; +}; + struct tetrahedral_17x17x17 { struct dc_rgb lut0[1229]; struct dc_rgb lut1[1228]; @@ -165,14 +172,23 @@ struct tetrahedral_9x9x9 { struct dc_rgb lut3[182]; }; +enum lut_dimension { + LUT_DIM_INVALID = 0, + LUT_DIM_9 = 9, + LUT_DIM_17 = 17, + LUT_DIM_33 = 33, +}; + struct tetrahedral_params { union { +//TODO: Uncomment when in use. +// struct tetrahedral_33x33x33 tetrahedral_33; struct tetrahedral_17x17x17 tetrahedral_17; struct tetrahedral_9x9x9 tetrahedral_9; }; bool use_tetrahedral_9; bool use_12bits; - + enum lut_dimension lut_dim; }; /* arr_curve_points - regamma regions/segments specification diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/mpc.h b/drivers/gpu/drm/amd/display/dc/inc/hw/mpc.h index 0db607f2a410..f5617674bea8 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/mpc.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/mpc.h @@ -54,6 +54,7 @@ #include "dc_hw_types.h" #include "hw_shared.h" #include "transform.h" +#include "dc_types.h" #define MAX_MPCC 6 #define MAX_OPP 6 @@ -66,7 +67,6 @@ enum mpc_output_csc_mode { MPC_OUTPUT_CSC_COEF_B }; - enum mpcc_blend_mode { MPCC_BLEND_MODE_BYPASS, MPCC_BLEND_MODE_TOP_LAYER_PASSTHROUGH, @@ -102,13 +102,6 @@ enum mpcc_movable_cm_location { MPCC_MOVABLE_CM_LOCATION_AFTER, }; -enum MCM_LUT_XABLE { - MCM_LUT_DISABLE, - MCM_LUT_DISABLED = MCM_LUT_DISABLE, - MCM_LUT_ENABLE, - MCM_LUT_ENABLED = MCM_LUT_ENABLE, -}; - enum MCM_LUT_ID { MCM_LUT_3DLUT, MCM_LUT_1DLUT, @@ -117,7 +110,7 @@ enum MCM_LUT_ID { struct mpc_fl_3dlut_config { bool enabled; - uint16_t width; + enum dc_cm_lut_size size; bool select_lut_bank_a; uint16_t bit_depth; int hubp_index; @@ -1042,22 +1035,22 @@ struct mpc_funcs { void (*update_3dlut_fast_load_select)(struct mpc *mpc, int mpcc_id, int hubp_idx); -/** - * @get_3dlut_fast_load_status: - * - * Get 3D LUT fast load status and reference them with done, soft_underflow and hard_underflow pointers. - * - * Parameters: - * - [in/out] mpc - MPC context. - * - [in] mpcc_id - * - [in/out] done - * - [in/out] soft_underflow - * - [in/out] hard_underflow - * - * Return: - * - * void - */ + /** + * @get_3dlut_fast_load_status: + * + * Get 3D LUT fast load status and reference them with done, soft_underflow and hard_underflow pointers. + * + * Parameters: + * - [in/out] mpc - MPC context. + * - [in] mpcc_id + * - [in/out] done + * - [in/out] soft_underflow + * - [in/out] hard_underflow + * + * Return: + * + * void + */ void (*get_3dlut_fast_load_status)(struct mpc *mpc, int mpcc_id, uint32_t *done, uint32_t *soft_underflow, uint32_t *hard_underflow); /** @@ -1076,8 +1069,11 @@ struct mpc_funcs { * * void */ - void (*populate_lut)(struct mpc *mpc, const enum MCM_LUT_ID id, const union mcm_lut_params params, - bool lut_bank_a, int mpcc_id); + void (*populate_lut)(struct mpc *mpc, + const enum MCM_LUT_ID id, + const union mcm_lut_params *params, + const bool lut_bank_a, + const int mpcc_id); /** * @program_lut_read_write_control: @@ -1088,13 +1084,18 @@ struct mpc_funcs { * - [in/out] mpc - MPC context. * - [in] id * - [in] lut_bank_a + * - [in] bit_depth * - [in] mpcc_id * * Return: * * void */ - void (*program_lut_read_write_control)(struct mpc *mpc, const enum MCM_LUT_ID id, bool lut_bank_a, int mpcc_id); + void (*program_lut_read_write_control)(struct mpc *mpc, + const enum MCM_LUT_ID id, + const bool lut_bank_a, + const unsigned int bit_depth, + const int mpcc_id); /** * @program_lut_mode: @@ -1104,33 +1105,44 @@ struct mpc_funcs { * Parameters: * - [in/out] mpc - MPC context. * - [in] id - * - [in] xable + * - [in] enable * - [in] lut_bank_a + * - [in] size * - [in] mpcc_id * * Return: * * void */ - void (*program_lut_mode)(struct mpc *mpc, const enum MCM_LUT_ID id, const enum MCM_LUT_XABLE xable, - bool lut_bank_a, int mpcc_id); + void (*program_lut_mode)(struct mpc *mpc, + const enum MCM_LUT_ID id, + const bool enable, + const bool lut_bank_a, + const enum dc_cm_lut_size size, + const int mpcc_id); - /** - * @mcm: - * - * MPC MCM new HW sequential programming functions - */ - struct { - void (*program_3dlut_size)(struct mpc *mpc, uint32_t width, int mpcc_id); - void (*program_bias_scale)(struct mpc *mpc, uint16_t bias, uint16_t scale, int mpcc_id); - void (*program_bit_depth)(struct mpc *mpc, uint16_t bit_depth, int mpcc_id); - bool (*is_config_supported)(uint32_t width); - void (*program_lut_read_write_control)(struct mpc *mpc, const enum MCM_LUT_ID id, - bool lut_bank_a, bool enabled, int mpcc_id); - void (*populate_lut)(struct mpc *mpc, const union mcm_lut_params params, - bool lut_bank_a, int mpcc_id); - } mcm; + /** + * @get_lut_mode: + * + * Obtains enablement and ram bank status. + * + * Parameters: + * - [in/out] mpc - MPC context. + * - [in] id + * - [in] mpcc_id + * - [out] enable + * - [out] lut_bank_a + * + * Return: + * + * void + */ + void (*get_lut_mode)(struct mpc *mpc, + const enum MCM_LUT_ID id, + const int mpcc_id, + bool *enable, + bool *lut_bank_a); /** * @rmcm: @@ -1143,9 +1155,11 @@ struct mpc_funcs { void (*update_3dlut_fast_load_select)(struct mpc *mpc, int mpcc_id, int hubp_idx); void (*program_lut_read_write_control)(struct mpc *mpc, const enum MCM_LUT_ID id, bool lut_bank_a, bool enabled, int mpcc_id); - void (*program_lut_mode)(struct mpc *mpc, const enum MCM_LUT_XABLE xable, - bool lut_bank_a, int mpcc_id); - void (*program_3dlut_size)(struct mpc *mpc, uint32_t width, int mpcc_id); + void (*program_lut_mode)(struct mpc *mpc, + bool enable, + bool lut_bank_a, + int mpcc_id); + void (*program_3dlut_size)(struct mpc *mpc, const enum dc_cm_lut_size size, int mpcc_id); void (*program_bias_scale)(struct mpc *mpc, uint16_t bias, uint16_t scale, int mpcc_id); void (*program_bit_depth)(struct mpc *mpc, uint16_t bit_depth, int mpcc_id); bool (*is_config_supported)(uint32_t width); diff --git a/drivers/gpu/drm/amd/display/dc/inc/resource.h b/drivers/gpu/drm/amd/display/dc/inc/resource.h index 79746d931471..cecd3282a29f 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/resource.h +++ b/drivers/gpu/drm/amd/display/dc/inc/resource.h @@ -60,6 +60,7 @@ struct resource_caps { int num_hpo_dp_stream_encoder; int num_hpo_dp_link_encoder; int num_mpc_3dlut; + int num_rmcm; }; struct resource_straps { diff --git a/drivers/gpu/drm/amd/display/dc/mpc/dcn401/dcn401_mpc.c b/drivers/gpu/drm/amd/display/dc/mpc/dcn401/dcn401_mpc.c index ce1ee2062e41..e0617db2d0c1 100644 --- a/drivers/gpu/drm/amd/display/dc/mpc/dcn401/dcn401_mpc.c +++ b/drivers/gpu/drm/amd/display/dc/mpc/dcn401/dcn401_mpc.c @@ -73,56 +73,15 @@ void mpc401_set_movable_cm_location(struct mpc *mpc, enum mpcc_movable_cm_locati } } -static enum dc_lut_mode get3dlut_config( - struct mpc *mpc, - bool *is_17x17x17, - bool *is_12bits_color_channel, - int mpcc_id) -{ - uint32_t i_mode, i_enable_10bits, lut_size; - enum dc_lut_mode mode; - struct dcn401_mpc *mpc401 = TO_DCN401_MPC(mpc); - - REG_GET(MPCC_MCM_3DLUT_MODE[mpcc_id], - MPCC_MCM_3DLUT_MODE_CURRENT, &i_mode); - - REG_GET(MPCC_MCM_3DLUT_READ_WRITE_CONTROL[mpcc_id], - MPCC_MCM_3DLUT_30BIT_EN, &i_enable_10bits); - - switch (i_mode) { - case 0: - mode = LUT_BYPASS; - break; - case 1: - mode = LUT_RAM_A; - break; - case 2: - mode = LUT_RAM_B; - break; - default: - mode = LUT_BYPASS; - break; - } - if (i_enable_10bits > 0) - *is_12bits_color_channel = false; - else - *is_12bits_color_channel = true; - - REG_GET(MPCC_MCM_3DLUT_MODE[mpcc_id], MPCC_MCM_3DLUT_SIZE, &lut_size); - - if (lut_size == 0) - *is_17x17x17 = true; - else - *is_17x17x17 = false; - - return mode; -} - -void mpc401_populate_lut(struct mpc *mpc, const enum MCM_LUT_ID id, const union mcm_lut_params params, bool lut_bank_a, int mpcc_id) +void mpc401_populate_lut(struct mpc *mpc, + const enum MCM_LUT_ID id, + const union mcm_lut_params *params, + const bool lut_bank_a, + const int mpcc_id) { const enum dc_lut_mode next_mode = lut_bank_a ? LUT_RAM_A : LUT_RAM_B; - const struct pwl_params *lut1d = params.pwl; - const struct pwl_params *lut_shaper = params.pwl; + const struct pwl_params *lut1d = params->pwl; + const struct pwl_params *lut_shaper = params->pwl; bool is_17x17x17; bool is_12bits_color_channel; const struct dc_rgb *lut0; @@ -131,7 +90,7 @@ void mpc401_populate_lut(struct mpc *mpc, const enum MCM_LUT_ID id, const union const struct dc_rgb *lut3; int lut_size0; int lut_size; - const struct tetrahedral_params *lut3d = params.lut3d; + const struct tetrahedral_params *lut3d = params->lut3d; switch (id) { case MCM_LUT_1DLUT: @@ -174,8 +133,6 @@ void mpc401_populate_lut(struct mpc *mpc, const enum MCM_LUT_ID id, const union mpc32_power_on_shaper_3dlut(mpc, mpcc_id, true); - get3dlut_config(mpc, &is_17x17x17, &is_12bits_color_channel, mpcc_id); - is_17x17x17 = !lut3d->use_tetrahedral_9; is_12bits_color_channel = lut3d->use_12bits; if (is_17x17x17) { @@ -198,8 +155,6 @@ void mpc401_populate_lut(struct mpc *mpc, const enum MCM_LUT_ID id, const union sizeof(lut3d->tetrahedral_9.lut1[0]); } - mpc32_select_3dlut_ram(mpc, next_mode, - is_12bits_color_channel, mpcc_id); mpc32_select_3dlut_ram_mask(mpc, 0x1, mpcc_id); if (is_12bits_color_channel) mpc32_set3dlut_ram12(mpc, lut0, lut_size0, mpcc_id); @@ -232,46 +187,69 @@ void mpc401_populate_lut(struct mpc *mpc, const enum MCM_LUT_ID id, const union } +static uint32_t mpc401_cm_lut_size_to_3dlut_size(const enum dc_cm_lut_size cm_size) +{ + uint32_t size = 0; + + switch (cm_size) { + case CM_LUT_SIZE_999: + size = 1; + break; + case CM_LUT_SIZE_171717: + size = 0; + break; + default: + /* invalid LUT size */ + ASSERT(false); + size = 0; + break; + } + + return size; +} + void mpc401_program_lut_mode( struct mpc *mpc, const enum MCM_LUT_ID id, - const enum MCM_LUT_XABLE xable, - bool lut_bank_a, - int mpcc_id) + const bool enable, + const bool lut_bank_a, + const enum dc_cm_lut_size size, + const int mpcc_id) { + uint32_t lut_size; struct dcn401_mpc *mpc401 = TO_DCN401_MPC(mpc); switch (id) { case MCM_LUT_3DLUT: - switch (xable) { - case MCM_LUT_DISABLE: + if (enable) { + lut_size = mpc401_cm_lut_size_to_3dlut_size(size); + REG_UPDATE_2(MPCC_MCM_3DLUT_MODE[mpcc_id], + MPCC_MCM_3DLUT_MODE, lut_bank_a ? 1 : 2, + MPCC_MCM_3DLUT_SIZE, lut_size); + } else { + if (mpc->ctx->dc->debug.enable_mem_low_power.bits.mpc) + mpc32_power_on_shaper_3dlut(mpc, mpcc_id, false); REG_UPDATE(MPCC_MCM_3DLUT_MODE[mpcc_id], MPCC_MCM_3DLUT_MODE, 0); - break; - case MCM_LUT_ENABLE: - REG_UPDATE(MPCC_MCM_3DLUT_MODE[mpcc_id], MPCC_MCM_3DLUT_MODE, lut_bank_a ? 1 : 2); - break; } break; case MCM_LUT_SHAPER: - switch (xable) { - case MCM_LUT_DISABLE: - REG_UPDATE(MPCC_MCM_SHAPER_CONTROL[mpcc_id], MPCC_MCM_SHAPER_LUT_MODE, 0); - break; - case MCM_LUT_ENABLE: + if (enable) { REG_UPDATE(MPCC_MCM_SHAPER_CONTROL[mpcc_id], MPCC_MCM_SHAPER_LUT_MODE, lut_bank_a ? 1 : 2); - break; + } else { + if (mpc->ctx->dc->debug.enable_mem_low_power.bits.mpc) + mpc32_power_on_shaper_3dlut(mpc, mpcc_id, false); + REG_UPDATE(MPCC_MCM_SHAPER_CONTROL[mpcc_id], MPCC_MCM_SHAPER_LUT_MODE, 0); } break; case MCM_LUT_1DLUT: - switch (xable) { - case MCM_LUT_DISABLE: - REG_UPDATE(MPCC_MCM_1DLUT_CONTROL[mpcc_id], - MPCC_MCM_1DLUT_MODE, 0); - break; - case MCM_LUT_ENABLE: + if (enable) { REG_UPDATE(MPCC_MCM_1DLUT_CONTROL[mpcc_id], MPCC_MCM_1DLUT_MODE, 2); - break; + } else { + if (mpc->ctx->dc->debug.enable_mem_low_power.bits.mpc) + mpc32_power_on_blnd_lut(mpc, mpcc_id, false); + REG_UPDATE(MPCC_MCM_1DLUT_CONTROL[mpcc_id], + MPCC_MCM_1DLUT_MODE, 0); } REG_UPDATE(MPCC_MCM_1DLUT_CONTROL[mpcc_id], MPCC_MCM_1DLUT_SELECT, lut_bank_a ? 0 : 1); @@ -279,14 +257,20 @@ void mpc401_program_lut_mode( } } -void mpc401_program_lut_read_write_control(struct mpc *mpc, const enum MCM_LUT_ID id, bool lut_bank_a, int mpcc_id) +void mpc401_program_lut_read_write_control(struct mpc *mpc, + const enum MCM_LUT_ID id, + const bool lut_bank_a, + const unsigned int bit_depth, + const int mpcc_id) { struct dcn401_mpc *mpc401 = TO_DCN401_MPC(mpc); switch (id) { case MCM_LUT_3DLUT: mpc32_select_3dlut_ram_mask(mpc, 0xf, mpcc_id); - REG_UPDATE(MPCC_MCM_3DLUT_READ_WRITE_CONTROL[mpcc_id], MPCC_MCM_3DLUT_RAM_SEL, lut_bank_a ? 0 : 1); + REG_UPDATE_2(MPCC_MCM_3DLUT_READ_WRITE_CONTROL[mpcc_id], + MPCC_MCM_3DLUT_30BIT_EN, (bit_depth == 10) ? 1 : 0, + MPCC_MCM_3DLUT_RAM_SEL, lut_bank_a ? 0 : 1); break; case MCM_LUT_SHAPER: mpc32_configure_shaper_lut(mpc, lut_bank_a, mpcc_id); @@ -578,6 +562,44 @@ void mpc401_get_gamut_remap(struct mpc *mpc, arr_reg_val, ARRAY_SIZE(arr_reg_val)); } +void mpc401_get_lut_mode(struct mpc *mpc, + const enum MCM_LUT_ID id, + const int mpcc_id, + bool *enable, + bool *lut_bank_a) +{ + struct dcn401_mpc *mpc401 = TO_DCN401_MPC(mpc); + + uint32_t lut_mode = 0; + uint32_t lut_select = 0; + + *enable = false; + *lut_bank_a = true; + + switch (id) { + case MCM_LUT_SHAPER: + REG_GET(MPCC_MCM_SHAPER_CONTROL[mpcc_id], + MPCC_MCM_SHAPER_MODE_CURRENT, &lut_mode); + *enable = lut_mode != 0; + *lut_bank_a = lut_mode != 2; + break; + case MCM_LUT_1DLUT: + REG_GET_2(MPCC_MCM_1DLUT_CONTROL[mpcc_id], + MPCC_MCM_1DLUT_MODE_CURRENT, &lut_mode, + MPCC_MCM_1DLUT_SELECT_CURRENT, &lut_select); + *enable = lut_mode != 0; + *lut_bank_a = lut_mode == 0 || lut_select == 0; + break; + case MCM_LUT_3DLUT: + default: + REG_GET(MPCC_MCM_3DLUT_MODE[mpcc_id], + MPCC_MCM_3DLUT_MODE_CURRENT, &lut_mode); + *enable = lut_mode != 0; + *lut_bank_a = lut_mode != 2; + break; + } +} + static const struct mpc_funcs dcn401_mpc_funcs = { .read_mpcc_state = mpc1_read_mpcc_state, .insert_plane = mpc1_insert_plane, @@ -616,6 +638,7 @@ static const struct mpc_funcs dcn401_mpc_funcs = { .populate_lut = mpc401_populate_lut, .program_lut_read_write_control = mpc401_program_lut_read_write_control, .program_lut_mode = mpc401_program_lut_mode, + .get_lut_mode = mpc401_get_lut_mode, }; diff --git a/drivers/gpu/drm/amd/display/dc/mpc/dcn401/dcn401_mpc.h b/drivers/gpu/drm/amd/display/dc/mpc/dcn401/dcn401_mpc.h index 6d842d7b95c7..c16560c84453 100644 --- a/drivers/gpu/drm/amd/display/dc/mpc/dcn401/dcn401_mpc.h +++ b/drivers/gpu/drm/amd/display/dc/mpc/dcn401/dcn401_mpc.h @@ -206,21 +206,32 @@ void dcn401_mpc_construct(struct dcn401_mpc *mpc401, int num_rmu); void mpc401_set_movable_cm_location(struct mpc *mpc, enum mpcc_movable_cm_location location, int mpcc_id); -void mpc401_populate_lut(struct mpc *mpc, const enum MCM_LUT_ID id, const union mcm_lut_params params, - bool lut_bank_a, int mpcc_id); +void mpc401_populate_lut(struct mpc *mpc, + const enum MCM_LUT_ID id, + const union mcm_lut_params *params, + bool lut_bank_a, + int mpcc_id); void mpc401_program_lut_mode( struct mpc *mpc, const enum MCM_LUT_ID id, - const enum MCM_LUT_XABLE xable, - bool lut_bank_a, - int mpcc_id); + const bool enable, + const bool lut_bank_a, + const enum dc_cm_lut_size size, + const int mpcc_id); + +void mpc401_get_lut_mode(struct mpc *mpc, + const enum MCM_LUT_ID id, + const int mpcc_id, + bool *enable, + bool *lut_bank_a); void mpc401_program_lut_read_write_control( struct mpc *mpc, const enum MCM_LUT_ID id, - bool lut_bank_a, - int mpcc_id); + const bool lut_bank_a, + const unsigned int bit_depth, + const int mpcc_id); void mpc401_set_gamut_remap( struct mpc *mpc, diff --git a/drivers/gpu/drm/amd/display/dc/mpc/dcn42/dcn42_mpc.c b/drivers/gpu/drm/amd/display/dc/mpc/dcn42/dcn42_mpc.c index 304b23109fb0..507dbdbea600 100644 --- a/drivers/gpu/drm/amd/display/dc/mpc/dcn42/dcn42_mpc.c +++ b/drivers/gpu/drm/amd/display/dc/mpc/dcn42/dcn42_mpc.c @@ -20,7 +20,7 @@ mpc42->mpc_shift->field_name, mpc42->mpc_mask->field_name -static void mpc42_init_mpcc(struct mpcc *mpcc, int mpcc_inst) +void mpc42_init_mpcc(struct mpcc *mpcc, int mpcc_inst) { mpcc->mpcc_id = mpcc_inst; mpcc->dpp_id = 0xf; @@ -63,154 +63,6 @@ void mpc42_update_blending( mpcc->blnd_cfg = *blnd_cfg; } -/* Shaper functions */ -void mpc42_power_on_shaper_3dlut( - struct mpc *mpc, - uint32_t mpcc_id, - bool power_on) -{ - uint32_t power_status_shaper = 2; - uint32_t power_status_3dlut = 2; - struct dcn42_mpc *mpc42 = TO_DCN42_MPC(mpc); - int max_retries = 10; - - REG_SET(MPCC_MCM_MEM_PWR_CTRL[mpcc_id], 0, - MPCC_MCM_3DLUT_MEM_PWR_DIS, power_on == true ? 1:0); - REG_SET(MPCC_MCM_MEM_PWR_CTRL[mpcc_id], 0, - MPCC_MCM_SHAPER_MEM_PWR_DIS, power_on == true ? 1:0); - /* wait for memory to fully power up */ - if (power_on && mpc->ctx->dc->debug.enable_mem_low_power.bits.mpc) { - REG_WAIT(MPCC_MCM_MEM_PWR_CTRL[mpcc_id], MPCC_MCM_SHAPER_MEM_PWR_STATE, 0, 1, max_retries); - REG_WAIT(MPCC_MCM_MEM_PWR_CTRL[mpcc_id], MPCC_MCM_3DLUT_MEM_PWR_STATE, 0, 1, max_retries); - } - - /*read status is not mandatory, it is just for debugging*/ - REG_GET(MPCC_MCM_MEM_PWR_CTRL[mpcc_id], MPCC_MCM_SHAPER_MEM_PWR_STATE, &power_status_shaper); - REG_GET(MPCC_MCM_MEM_PWR_CTRL[mpcc_id], MPCC_MCM_3DLUT_MEM_PWR_STATE, &power_status_3dlut); - - if (power_status_shaper != 0 && power_on == true) - BREAK_TO_DEBUGGER(); - - if (power_status_3dlut != 0 && power_on == true) - BREAK_TO_DEBUGGER(); -} - -void mpc42_configure_shaper_lut( - struct mpc *mpc, - bool is_ram_a, - uint32_t mpcc_id) -{ - struct dcn42_mpc *mpc42 = TO_DCN42_MPC(mpc); - - REG_UPDATE(MPCC_MCM_SHAPER_SCALE_G_B[mpcc_id], - MPCC_MCM_SHAPER_SCALE_B, 0x7000); - REG_UPDATE(MPCC_MCM_SHAPER_SCALE_G_B[mpcc_id], - MPCC_MCM_SHAPER_SCALE_G, 0x7000); - REG_UPDATE(MPCC_MCM_SHAPER_SCALE_R[mpcc_id], - MPCC_MCM_SHAPER_SCALE_R, 0x7000); - REG_UPDATE(MPCC_MCM_SHAPER_LUT_WRITE_EN_MASK[mpcc_id], - MPCC_MCM_SHAPER_LUT_WRITE_EN_MASK, 7); - REG_UPDATE(MPCC_MCM_SHAPER_LUT_WRITE_EN_MASK[mpcc_id], - MPCC_MCM_SHAPER_LUT_WRITE_SEL, is_ram_a == true ? 0:1); - REG_SET(MPCC_MCM_SHAPER_LUT_INDEX[mpcc_id], 0, MPCC_MCM_SHAPER_LUT_INDEX, 0); -} - - -void mpc42_program_3dlut_size(struct mpc *mpc, uint32_t width, int mpcc_id) -{ - struct dcn42_mpc *mpc42 = TO_DCN42_MPC(mpc); - uint32_t size = 0xff; - - REG_GET(MPCC_MCM_3DLUT_MODE[mpcc_id], MPCC_MCM_3DLUT_SIZE, &size); - - REG_UPDATE(MPCC_MCM_3DLUT_MODE[mpcc_id], MPCC_MCM_3DLUT_SIZE, - (width == 33) ? 2 : - (width == 17) ? 0 : 2); - - REG_GET(MPCC_MCM_3DLUT_MODE[mpcc_id], MPCC_MCM_3DLUT_SIZE, &size); -} - -void mpc42_program_3dlut_fl_bias_scale(struct mpc *mpc, uint16_t bias, uint16_t scale, int mpcc_id) -{ - struct dcn42_mpc *mpc42 = TO_DCN42_MPC(mpc); - - REG_UPDATE_2(MPCC_MCM_3DLUT_OUT_OFFSET_R[mpcc_id], - MPCC_MCM_3DLUT_OUT_OFFSET_R, bias, - MPCC_MCM_3DLUT_OUT_SCALE_R, scale); - - REG_UPDATE_2(MPCC_MCM_3DLUT_OUT_OFFSET_G[mpcc_id], - MPCC_MCM_3DLUT_OUT_OFFSET_G, bias, - MPCC_MCM_3DLUT_OUT_SCALE_G, scale); - - REG_UPDATE_2(MPCC_MCM_3DLUT_OUT_OFFSET_B[mpcc_id], - MPCC_MCM_3DLUT_OUT_OFFSET_B, bias, - MPCC_MCM_3DLUT_OUT_SCALE_B, scale); -} - -void mpc42_program_bit_depth(struct mpc *mpc, uint16_t bit_depth, int mpcc_id) -{ - struct dcn42_mpc *mpc42 = TO_DCN42_MPC(mpc); - - REG_UPDATE(MPCC_MCM_3DLUT_READ_WRITE_CONTROL[mpcc_id], MPCC_MCM_3DLUT_WRITE_EN_MASK, 0xF); - - //program bit_depth - REG_UPDATE(MPCC_MCM_3DLUT_READ_WRITE_CONTROL[mpcc_id], - MPCC_MCM_3DLUT_30BIT_EN, - (bit_depth == 10) ? 1 : 0); -} - -bool mpc42_is_config_supported(uint32_t width) -{ - if (width == 17) - return true; - - return false; -} - -void mpc42_populate_lut(struct mpc *mpc, const union mcm_lut_params params, - bool lut_bank_a, int mpcc_id) -{ - const enum dc_lut_mode next_mode = lut_bank_a ? LUT_RAM_A : LUT_RAM_B; - const struct pwl_params *lut_shaper = params.pwl; - - if (lut_shaper == NULL) - return; - if (mpc->ctx->dc->debug.enable_mem_low_power.bits.mpc) - mpc42_power_on_shaper_3dlut(mpc, mpcc_id, true); - - mpc42_configure_shaper_lut(mpc, next_mode == LUT_RAM_A, mpcc_id); - - if (next_mode == LUT_RAM_A) - mpc32_program_shaper_luta_settings(mpc, lut_shaper, mpcc_id); - else - mpc32_program_shaper_lutb_settings(mpc, lut_shaper, mpcc_id); - - mpc32_program_shaper_lut( - mpc, lut_shaper->rgb_resulted, lut_shaper->hw_points_num, mpcc_id); - - mpc42_power_on_shaper_3dlut(mpc, mpcc_id, false); -} - -void mpc42_program_lut_read_write_control(struct mpc *mpc, const enum MCM_LUT_ID id, - bool lut_bank_a, bool enabled, int mpcc_id) -{ - struct dcn42_mpc *mpc42 = TO_DCN42_MPC(mpc); - - switch (id) { - case MCM_LUT_3DLUT: - REG_UPDATE(MPCC_MCM_3DLUT_MODE[mpcc_id], MPCC_MCM_3DLUT_MODE, - (!enabled) ? 0 : - (lut_bank_a) ? 1 : 2); - REG_UPDATE(MPCC_MCM_3DLUT_READ_WRITE_CONTROL[mpcc_id], MPCC_MCM_3DLUT_RAM_SEL, lut_bank_a ? 0 : 1); - break; - case MCM_LUT_SHAPER: - mpc32_configure_shaper_lut(mpc, lut_bank_a, mpcc_id); - break; - default: - break; - } -} - /* RMCM Shaper functions */ void mpc42_power_on_rmcm_shaper_3dlut( struct mpc *mpc, @@ -674,32 +526,47 @@ void mpc42_program_rmcm_lut_read_write_control(struct mpc *mpc, const enum MCM_L } } -void mpc42_program_lut_mode(struct mpc *mpc, const enum MCM_LUT_XABLE xable, - bool lut_bank_a, int mpcc_id) +void mpc42_program_lut_mode(struct mpc *mpc, + bool enable, + bool lut_bank_a, + int mpcc_id) { struct dcn42_mpc *mpc42 = TO_DCN42_MPC(mpc); - switch (xable) { - case MCM_LUT_DISABLE: + if (enable) { + REG_UPDATE(MPC_RMCM_SHAPER_CONTROL[mpcc_id], MPC_RMCM_SHAPER_LUT_MODE, lut_bank_a ? 1 : 2); + } else { REG_UPDATE(MPC_RMCM_SHAPER_CONTROL[mpcc_id], MPC_RMCM_SHAPER_LUT_MODE, 0); + } +} + +static uint32_t mpc42_get_rmcm_3dlut_width( + const enum dc_cm_lut_size size) +{ + uint32_t width = 0; + + switch (size) { + case CM_LUT_SIZE_333333: + width = 2; break; - case MCM_LUT_ENABLE: - REG_UPDATE(MPC_RMCM_SHAPER_CONTROL[mpcc_id], MPC_RMCM_SHAPER_LUT_MODE, lut_bank_a ? 1 : 2); + case CM_LUT_SIZE_171717: + default: + width = 0; break; } + + return width; } -void mpc42_program_rmcm_3dlut_size(struct mpc *mpc, uint32_t width, int mpcc_id) +void mpc42_program_rmcm_3dlut_size(struct mpc *mpc, + const enum dc_cm_lut_size size, + int mpcc_id) { struct dcn42_mpc *mpc42 = TO_DCN42_MPC(mpc); - uint32_t size = 0xff; - - REG_GET(MPC_RMCM_3DLUT_MODE[mpcc_id], MPC_RMCM_3DLUT_SIZE, &size); + uint32_t width = mpc42_get_rmcm_3dlut_width(size); - REG_UPDATE(MPC_RMCM_3DLUT_MODE[mpcc_id], MPC_RMCM_3DLUT_SIZE, - (width == 33) ? 2 : 0); - - REG_GET(MPC_RMCM_3DLUT_MODE[mpcc_id], MPC_RMCM_3DLUT_SIZE, &size); + REG_UPDATE(MPC_RMCM_3DLUT_MODE[mpcc_id], + MPC_RMCM_3DLUT_SIZE, width); } void mpc42_program_rmcm_3dlut_fast_load_bias_scale(struct mpc *mpc, uint16_t bias, uint16_t scale, int mpcc_id) @@ -731,14 +598,6 @@ void mpc42_program_rmcm_bit_depth(struct mpc *mpc, uint16_t bit_depth, int mpcc_ (bit_depth == 10) ? 1 : 0); } -bool mpc42_is_rmcm_config_supported(uint32_t width) -{ - if (width == 17 || width == 33) - return true; - - return false; -} - void mpc42_set_fl_config( struct mpc *mpc, struct mpc_fl_3dlut_config *cfg, @@ -746,6 +605,7 @@ void mpc42_set_fl_config( { struct dcn42_mpc *mpc42 = TO_DCN42_MPC(mpc); + uint32_t width = mpc42_get_rmcm_3dlut_width(cfg->size); /* From: Jie Zhou @@ -786,7 +646,7 @@ void mpc42_set_fl_config( //width REG_UPDATE_2(MPC_RMCM_3DLUT_MODE[mpcc_id], - MPC_RMCM_3DLUT_SIZE, (cfg->width == 33) ? 2 : 0, + MPC_RMCM_3DLUT_SIZE, width, MPC_RMCM_3DLUT_MODE, (!cfg->enabled) ? 0 : (cfg->select_lut_bank_a) ? 1 : 2); //connect to hubp @@ -799,182 +659,6 @@ void mpc42_set_fl_config( REG_UPDATE(MPC_RMCM_CNTL[mpcc_id], MPC_RMCM_CNTL, cfg->enabled ? 0 : 0xF); } -//static void rmcm_program_gamut_remap( -// struct mpc *mpc, -// unsigned int mpcc_id, -// const uint16_t *regval, -// enum mpcc_gamut_remap_id gamut_remap_block_id, -// enum mpcc_gamut_remap_mode_select mode_select) -//{ -// struct color_matrices_reg gamut_regs; -// struct dcn42_mpc *mpc42 = TO_DCN42_MPC(mpc); -// -// if (gamut_remap_block_id == MPCC_OGAM_GAMUT_REMAP || -// gamut_remap_block_id == MPCC_MCM_FIRST_GAMUT_REMAP || -// gamut_remap_block_id == MPCC_MCM_SECOND_GAMUT_REMAP) { -// mpc_program_gamut_remap(mpc, mpcc_id, regval, gamut_remap_block_id, mode_select); -// return; -// } -// if (gamut_remap_block_id == MPCC_OGAM_GAMUT_REMAP) { -// -// if (regval == NULL || mode_select == MPCC_GAMUT_REMAP_MODE_SELECT_0) { -// REG_SET(MPC_RMCM_GAMUT_REMAP_MODE[mpcc_id], 0, -// MPC_RMCM_GAMUT_REMAP_MODE, mode_select); -// return; -// } -// -// gamut_regs.shifts.csc_c11 = mpc42->mpc_shift->MPCC_GAMUT_REMAP_C11_A; -// gamut_regs.masks.csc_c11 = mpc42->mpc_mask->MPCC_GAMUT_REMAP_C11_A; -// gamut_regs.shifts.csc_c12 = mpc42->mpc_shift->MPCC_GAMUT_REMAP_C12_A; -// gamut_regs.masks.csc_c12 = mpc42->mpc_mask->MPCC_GAMUT_REMAP_C12_A; -// -// switch (mode_select) { -// case MPCC_GAMUT_REMAP_MODE_SELECT_1: -// gamut_regs.csc_c11_c12 = REG(MPC_RMCM_GAMUT_REMAP_C11_C12_A[mpcc_id]); -// gamut_regs.csc_c33_c34 = REG(MPC_RMCM_GAMUT_REMAP_C33_C34_A[mpcc_id]); -// break; -// case MPCC_GAMUT_REMAP_MODE_SELECT_2: -// gamut_regs.csc_c11_c12 = REG(MPC_RMCM_GAMUT_REMAP_C11_C12_B[mpcc_id]); -// gamut_regs.csc_c33_c34 = REG(MPC_RMCM_GAMUT_REMAP_C33_C34_B[mpcc_id]); -// break; -// default: -// break; -// } -// -// cm_helper_program_color_matrices( -// mpc->ctx, -// regval, -// &gamut_regs); -// -// //select coefficient set to use, set A (MODE_1) or set B (MODE_2) -// REG_SET(MPC_RMCM_GAMUT_REMAP_MODE[mpcc_id], 0, MPC_RMCM_GAMUT_REMAP_MODE, mode_select); -// } -//} - -//static bool is_mpc_legacy_gamut_id(enum mpcc_gamut_remap_id gamut_remap_block_id) -//{ -// if (gamut_remap_block_id == MPCC_OGAM_GAMUT_REMAP || -// gamut_remap_block_id == MPCC_MCM_FIRST_GAMUT_REMAP || -// gamut_remap_block_id == MPCC_MCM_SECOND_GAMUT_REMAP) { -// return true; -// } -// return false; -//} -//static void program_gamut_remap( -// struct mpc *mpc, -// unsigned int mpcc_id, -// const uint16_t *regval, -// enum mpcc_gamut_remap_id gamut_remap_block_id, -// enum mpcc_gamut_remap_mode_select mode_select) -//{ -// if (is_mpc_legacy_gamut_id(gamut_remap_block_id)) -// mpc_program_gamut_remap(mpc, mpcc_id, regval, gamut_remap_block_id, mode_select); -// else -// rmcm_program_gamut_remap(mpc, mpcc_id, regval, gamut_remap_block_id, mode_select); -//} - -//void mpc42_set_gamut_remap( -// struct mpc *mpc, -// int mpcc_id, -// const struct mpc_grph_gamut_adjustment *adjust) -//{ -// struct dcn42_mpc *mpc42 = TO_DCN42_MPC(mpc); -// unsigned int i = 0; -// uint32_t mode_select = 0; -// -// if (adjust->gamut_adjust_type != GRAPHICS_GAMUT_ADJUST_TYPE_SW) { -// /* Bypass / Disable if type is bypass or hw */ -// program_gamut_remap(mpc, mpcc_id, NULL, -// adjust->mpcc_gamut_remap_block_id, MPCC_GAMUT_REMAP_MODE_SELECT_0); -// } else { -// struct fixed31_32 arr_matrix[12]; -// uint16_t arr_reg_val[12]; -// -// for (i = 0; i < 12; i++) -// arr_matrix[i] = adjust->temperature_matrix[i]; -// -// convert_float_matrix(arr_reg_val, arr_matrix, 12); -// -// if (is_mpc_legacy_gamut_id(adjust->mpcc_gamut_remap_block_id)) -// REG_GET(MPCC_GAMUT_REMAP_MODE[mpcc_id], -// MPCC_GAMUT_REMAP_MODE_CURRENT, &mode_select); -// else -// REG_GET(MPC_RMCM_GAMUT_REMAP_MODE[mpcc_id], -// MPC_RMCM_GAMUT_REMAP_MODE_CURRENT, &mode_select); -// -// //If current set in use not set A (MODE_1), then use set A, otherwise use set B -// if (mode_select != MPCC_GAMUT_REMAP_MODE_SELECT_1) -// mode_select = MPCC_GAMUT_REMAP_MODE_SELECT_1; -// else -// mode_select = MPCC_GAMUT_REMAP_MODE_SELECT_2; -// -// program_gamut_remap(mpc, mpcc_id, arr_reg_val, -// adjust->mpcc_gamut_remap_block_id, mode_select); -// } -//} - -//static void read_gamut_remap(struct mpc *mpc, -// int mpcc_id, -// uint16_t *regval, -// enum mpcc_gamut_remap_id gamut_remap_block_id, -// uint32_t *mode_select) -//{ -// struct color_matrices_reg gamut_regs = {0}; -// struct dcn42_mpc *mpc42 = TO_DCN42_MPC(mpc); -// -// if (is_mpc_legacy_gamut_id(gamut_remap_block_id)) { -// mpc_read_gamut_remap(mpc, mpcc_id, regval, gamut_remap_block_id, mode_select); -// } -// if (gamut_remap_block_id == MPCC_RMCM_GAMUT_REMAP) { -// //current coefficient set in use -// REG_GET(MPC_RMCM_GAMUT_REMAP_MODE[mpcc_id], MPC_RMCM_GAMUT_REMAP_MODE, mode_select); -// -// gamut_regs.shifts.csc_c11 = mpc42->mpc_shift->MPCC_GAMUT_REMAP_C11_A; -// gamut_regs.masks.csc_c11 = mpc42->mpc_mask->MPCC_GAMUT_REMAP_C11_A; -// gamut_regs.shifts.csc_c12 = mpc42->mpc_shift->MPCC_GAMUT_REMAP_C12_A; -// gamut_regs.masks.csc_c12 = mpc42->mpc_mask->MPCC_GAMUT_REMAP_C12_A; -// -// switch (*mode_select) { -// case MPCC_GAMUT_REMAP_MODE_SELECT_1: -// gamut_regs.csc_c11_c12 = REG(MPC_RMCM_GAMUT_REMAP_C11_C12_A[mpcc_id]); -// gamut_regs.csc_c33_c34 = REG(MPC_RMCM_GAMUT_REMAP_C33_C34_A[mpcc_id]); -// break; -// case MPCC_GAMUT_REMAP_MODE_SELECT_2: -// gamut_regs.csc_c11_c12 = REG(MPC_RMCM_GAMUT_REMAP_C11_C12_B[mpcc_id]); -// gamut_regs.csc_c33_c34 = REG(MPC_RMCM_GAMUT_REMAP_C33_C34_B[mpcc_id]); -// break; -// default: -// break; -// } -// } -// -// if (*mode_select != MPCC_GAMUT_REMAP_MODE_SELECT_0) { -// cm_helper_read_color_matrices( -// mpc42->base.ctx, -// regval, -// &gamut_regs); -// } -//} - -//void mpc42_get_gamut_remap(struct mpc *mpc, -// int mpcc_id, -// struct mpc_grph_gamut_adjustment *adjust) -//{ -// uint16_t arr_reg_val[12] = {0}; -// uint32_t mode_select; -// -// read_gamut_remap(mpc, mpcc_id, arr_reg_val, adjust->mpcc_gamut_remap_block_id, &mode_select); -// -// if (mode_select == MPCC_GAMUT_REMAP_MODE_SELECT_0) { -// adjust->gamut_adjust_type = GRAPHICS_GAMUT_ADJUST_TYPE_BYPASS; -// return; -// } -// -// adjust->gamut_adjust_type = GRAPHICS_GAMUT_ADJUST_TYPE_SW; -// convert_hw_matrix(adjust->temperature_matrix, -// arr_reg_val, ARRAY_SIZE(arr_reg_val)); -//} - void mpc42_read_mpcc_state( struct mpc *mpc, int mpcc_inst, @@ -1071,14 +755,7 @@ static const struct mpc_funcs dcn42_mpc_funcs = { .populate_lut = mpc401_populate_lut, .program_lut_read_write_control = mpc401_program_lut_read_write_control, .program_lut_mode = mpc401_program_lut_mode, - .mcm = { - .program_lut_read_write_control = mpc42_program_lut_read_write_control, - .program_3dlut_size = mpc42_program_3dlut_size, - .program_bias_scale = mpc42_program_3dlut_fl_bias_scale, - .program_bit_depth = mpc42_program_bit_depth, - .is_config_supported = mpc42_is_config_supported, - .populate_lut = mpc42_populate_lut, - }, + .get_lut_mode = mpc401_get_lut_mode, .rmcm = { .enable_3dlut_fl = mpc42_enable_3dlut_fl, .update_3dlut_fast_load_select = mpc42_update_3dlut_fast_load_select, @@ -1087,7 +764,6 @@ static const struct mpc_funcs dcn42_mpc_funcs = { .program_3dlut_size = mpc42_program_rmcm_3dlut_size, .program_bias_scale = mpc42_program_rmcm_3dlut_fast_load_bias_scale, .program_bit_depth = mpc42_program_rmcm_bit_depth, - .is_config_supported = mpc42_is_rmcm_config_supported, .power_on_shaper_3dlut = mpc42_power_on_rmcm_shaper_3dlut, .populate_lut = mpc42_populate_rmcm_lut, .fl_3dlut_configure = mpc42_set_fl_config, diff --git a/drivers/gpu/drm/amd/display/dc/mpc/dcn42/dcn42_mpc.h b/drivers/gpu/drm/amd/display/dc/mpc/dcn42/dcn42_mpc.h index 9b87fd2be904..a5f7f4f2bb3b 100644 --- a/drivers/gpu/drm/amd/display/dc/mpc/dcn42/dcn42_mpc.h +++ b/drivers/gpu/drm/amd/display/dc/mpc/dcn42/dcn42_mpc.h @@ -882,49 +882,7 @@ void dcn42_mpc_construct(struct dcn42_mpc *mpc401, int num_mpcc, int num_rmu); - -void mpc42_program_shaper_lutb_settings( - struct mpc *mpc, - const struct pwl_params *params, - uint32_t mpcc_id); -void mpc42_program_shaper_luta_settings( - struct mpc *mpc, - const struct pwl_params *params, - uint32_t mpcc_id); -void mpc42_configure_shaper_lut( - struct mpc *mpc, - bool is_ram_a, - uint32_t mpcc_id); -void mpc42_power_on_shaper_3dlut( - struct mpc *mpc, - uint32_t mpcc_id, - bool power_on); -void mpc42_program_3dlut_size( - struct mpc *mpc, - uint32_t width, - int mpcc_id); -void mpc42_program_3dlut_fl_bias_scale( - struct mpc *mpc, - uint16_t bias, - uint16_t scale, - int mpcc_id); -void mpc42_program_bit_depth( - struct mpc *mpc, - uint16_t bit_depth, - int mpcc_id); -void mpc42_populate_lut( - struct mpc *mpc, - const union mcm_lut_params params, - bool lut_bank_a, - int mpcc_id); -void mpc42_program_lut_read_write_control( - struct mpc *mpc, - const enum MCM_LUT_ID id, - bool lut_bank_a, - bool enabled, - int mpcc_id); - -bool mpc42_is_config_supported(uint32_t width); +void mpc42_init_mpcc(struct mpcc *mpcc, int mpcc_inst); /* RMCM */ void mpc42_program_rmcm_shaper_lut( @@ -969,12 +927,12 @@ void mpc42_program_rmcm_lut_read_write_control( int mpcc_id); void mpc42_program_lut_mode( struct mpc *mpc, - const enum MCM_LUT_XABLE xable, + bool enable, bool lut_bank_a, int mpcc_id); void mpc42_program_rmcm_3dlut_size( struct mpc *mpc, - uint32_t width, + const enum dc_cm_lut_size size, int mpcc_id); void mpc42_program_rmcm_3dlut_fast_load_bias_scale( struct mpc *mpc, @@ -986,8 +944,6 @@ void mpc42_program_rmcm_bit_depth( uint16_t bit_depth, int mpcc_id); -bool mpc42_is_rmcm_config_supported(uint32_t width); - void mpc42_set_fl_config( struct mpc *mpc, struct mpc_fl_3dlut_config *cfg, diff --git a/drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.h b/drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.h index d2ba63046a50..138081e6cc97 100644 --- a/drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.h +++ b/drivers/gpu/drm/amd/display/dc/optc/dcn10/dcn10_optc.h @@ -707,7 +707,7 @@ struct dcn_optc_shift { TG_REG_FIELD_LIST_DCN3_5(uint8_t) TG_REG_FIELD_LIST_DCN3_6(uint8_t) TG_REG_FIELD_LIST_DCN401(uint8_t) - TG_REG_FIELD_LIST_DCN42(uint8_t) + TG_REG_FIELD_LIST_DCN42(uint8_t) }; struct dcn_optc_mask { diff --git a/drivers/gpu/drm/amd/display/dc/optc/dcn42/dcn42_optc.c b/drivers/gpu/drm/amd/display/dc/optc/dcn42/dcn42_optc.c index effd05b3685f..ed66a2bbb8ae 100644 --- a/drivers/gpu/drm/amd/display/dc/optc/dcn42/dcn42_optc.c +++ b/drivers/gpu/drm/amd/display/dc/optc/dcn42/dcn42_optc.c @@ -6,11 +6,13 @@ #include "dcn30/dcn30_optc.h" #include "dcn31/dcn31_optc.h" #include "dcn32/dcn32_optc.h" +#include "dcn35/dcn35_optc.h" #include "dcn401/dcn401_optc.h" #include "reg_helper.h" #include "dc.h" #include "dcn_calc_math.h" #include "dc_dmub_srv.h" +#include "dc_trace.h" #define REG(reg)\ optc1->tg_regs->reg @@ -108,6 +110,89 @@ void optc42_disable_pwa(struct timing_generator *optc) REG_UPDATE(OTG_PWA_FRAME_SYNC_CONTROL, OTG_PWA_FRAME_SYNC_EN, 0); } +void optc42_clear_optc_underflow(struct timing_generator *optc) +{ + struct optc *optc1 = DCN10TG_FROM_TG(optc); + + REG_UPDATE(OPTC_INPUT_GLOBAL_CONTROL, OPTC_UNDERFLOW_CLEAR, 1); + REG_UPDATE(OPTC_RSMU_UNDERFLOW, OPTC_RSMU_UNDERFLOW_CLEAR, 1); +} +bool optc42_is_optc_underflow_occurred(struct timing_generator *optc) +{ + struct optc *optc1 = DCN10TG_FROM_TG(optc); + uint32_t underflow_occurred = 0, rsmu_underflow_occurred = 0; + + REG_GET(OPTC_INPUT_GLOBAL_CONTROL, + OPTC_UNDERFLOW_OCCURRED_STATUS, + &underflow_occurred); + + REG_GET(OPTC_RSMU_UNDERFLOW, + OPTC_RSMU_UNDERFLOW_OCCURRED_STATUS, + &rsmu_underflow_occurred); + return (underflow_occurred == 1 || rsmu_underflow_occurred); +} +/* disable_crtc */ +bool optc42_disable_crtc(struct timing_generator *optc) +{ + optc401_disable_crtc(optc); + optc42_clear_optc_underflow(optc); + + return true; +} +static void optc42_set_timing_double_buffer(struct timing_generator *optc, bool enable) +{ + struct optc *optc1 = DCN10TG_FROM_TG(optc); + uint32_t mode = enable ? 2 : 0; + /* actually we have 4 modes now, use as the same as previous dcn3x + * 00 OTG_DOUBLE_BUFFER_CONTROL_OTG_DRR_TIMING_DBUF_UPDATE_MODE_0 Double buffer update occurs at any time in a frame. + * 01 OTG_DOUBLE_BUFFER_CONTROL_OTG_DRR_TIMING_DBUF_UPDATE_MODE_1 Double buffer update occurs at OTG start of frame. + * 02 OTG_DOUBLE_BUFFER_CONTROL_OTG_DRR_TIMING_DBUF_UPDATE_MODE_2 Double buffer occurs DP start of frame. + * 03 OTG_DOUBLE_BUFFER_CONTROL_OTG_DRR_TIMING_DBUF_UPDATE_MODE_3 Reserved. + */ + + REG_UPDATE(OTG_DOUBLE_BUFFER_CONTROL, + OTG_DRR_TIMING_DBUF_UPDATE_MODE, mode); +} +void optc42_tg_init(struct timing_generator *optc) +{ + optc42_set_timing_double_buffer(optc, true); + optc42_clear_optc_underflow(optc); +} + +void optc42_lock_doublebuffer_enable(struct timing_generator *optc) +{ + struct optc *optc1 = DCN10TG_FROM_TG(optc); + uint32_t v_blank_start = 0; + uint32_t v_blank_end = 0; + uint32_t h_blank_start = 0; + uint32_t h_blank_end = 0; + + REG_GET_2(OTG_V_BLANK_START_END, + OTG_V_BLANK_START, &v_blank_start, + OTG_V_BLANK_END, &v_blank_end); + REG_GET_2(OTG_H_BLANK_START_END, + OTG_H_BLANK_START, &h_blank_start, + OTG_H_BLANK_END, &h_blank_end); + + REG_UPDATE_2(OTG_GLOBAL_CONTROL1, + MASTER_UPDATE_LOCK_DB_START_Y, v_blank_start, + MASTER_UPDATE_LOCK_DB_END_Y, v_blank_start); + REG_UPDATE_2(OTG_GLOBAL_CONTROL4, + DIG_UPDATE_POSITION_X, 20, + DIG_UPDATE_POSITION_Y, v_blank_start); + REG_UPDATE_3(OTG_GLOBAL_CONTROL0, + MASTER_UPDATE_LOCK_DB_START_X, h_blank_start - 200 - 1, + MASTER_UPDATE_LOCK_DB_END_X, h_blank_end, + MASTER_UPDATE_LOCK_DB_EN, 1); + REG_UPDATE(OTG_GLOBAL_CONTROL2, GLOBAL_UPDATE_LOCK_EN, 1); + + REG_SET_3(OTG_VUPDATE_KEEPOUT, 0, + MASTER_UPDATE_LOCK_VUPDATE_KEEPOUT_START_OFFSET, 0, + MASTER_UPDATE_LOCK_VUPDATE_KEEPOUT_END_OFFSET, 100, + OTG_MASTER_UPDATE_LOCK_VUPDATE_KEEPOUT_EN, 1); + + TRACE_OPTC_LOCK_UNLOCK_STATE(optc1, optc->inst, true); +} static struct timing_generator_funcs dcn42_tg_funcs = { .validate_timing = optc1_validate_timing, @@ -117,7 +202,7 @@ static struct timing_generator_funcs dcn42_tg_funcs = { .setup_vertical_interrupt2 = optc1_setup_vertical_interrupt2, .program_global_sync = optc401_program_global_sync, .enable_crtc = optc401_enable_crtc, - .disable_crtc = optc401_disable_crtc, + .disable_crtc = optc42_disable_crtc, .phantom_crtc_post_enable = optc401_phantom_crtc_post_enable, .disable_phantom_crtc = optc401_disable_phantom_otg, /* used by enable_timing_synchronization. Not need for FPGA */ @@ -138,7 +223,7 @@ static struct timing_generator_funcs dcn42_tg_funcs = { .disable_reset_trigger = optc1_disable_reset_trigger, .lock = optc3_lock, .unlock = optc1_unlock, - .lock_doublebuffer_enable = optc3_lock_doublebuffer_enable, + .lock_doublebuffer_enable = optc42_lock_doublebuffer_enable, .lock_doublebuffer_disable = optc3_lock_doublebuffer_disable, .enable_optc_clock = optc1_enable_optc_clock, .set_drr = optc401_set_drr, @@ -147,13 +232,13 @@ static struct timing_generator_funcs dcn42_tg_funcs = { .set_static_screen_control = optc1_set_static_screen_control, .program_stereo = optc1_program_stereo, .is_stereo_left_eye = optc1_is_stereo_left_eye, - .tg_init = optc3_tg_init, + .tg_init = optc42_tg_init, .is_tg_enabled = optc1_is_tg_enabled, - .is_optc_underflow_occurred = optc1_is_optc_underflow_occurred, - .clear_optc_underflow = optc1_clear_optc_underflow, + .is_optc_underflow_occurred = optc42_is_optc_underflow_occurred, + .clear_optc_underflow = optc42_clear_optc_underflow, .setup_global_swap_lock = NULL, .get_crc = optc42_get_crc, - .configure_crc = optc1_configure_crc, + .configure_crc = optc35_configure_crc, .set_dsc_config = optc3_set_dsc_config, .get_dsc_status = optc2_get_dsc_status, .set_dwb_source = NULL, @@ -162,6 +247,7 @@ static struct timing_generator_funcs dcn42_tg_funcs = { .wait_odm_doublebuffer_pending_clear = optc32_wait_odm_doublebuffer_pending_clear, .set_h_timing_div_manual_mode = optc401_set_h_timing_div_manual_mode, .get_optc_source = optc2_get_optc_source, + .wait_otg_disable = optc35_wait_otg_disable, .set_out_mux = optc401_set_out_mux, .set_drr_trigger_window = optc3_set_drr_trigger_window, .set_vtotal_change_limit = optc3_set_vtotal_change_limit, @@ -171,6 +257,8 @@ static struct timing_generator_funcs dcn42_tg_funcs = { .program_manual_trigger = optc2_program_manual_trigger, .setup_manual_trigger = optc2_setup_manual_trigger, .get_hw_timing = optc1_get_hw_timing, + .init_odm = optc3_init_odm, + .set_long_vtotal = optc35_set_long_vtotal, .is_two_pixels_per_container = optc1_is_two_pixels_per_container, .get_optc_double_buffer_pending = optc3_get_optc_double_buffer_pending, .get_otg_double_buffer_pending = optc3_get_otg_update_pending, @@ -178,6 +266,7 @@ static struct timing_generator_funcs dcn42_tg_funcs = { .set_vupdate_keepout = optc401_set_vupdate_keepout, .wait_update_lock_status = optc401_wait_update_lock_status, .optc_read_reg_state = optc31_read_reg_state, + .read_otg_state = optc31_read_otg_state, .enable_otg_pwa = optc42_enable_pwa, .disable_otg_pwa = optc42_disable_pwa, }; @@ -194,5 +283,9 @@ void dcn42_timing_generator_init(struct optc *optc1) optc1->min_v_blank_interlace = 5; optc1->min_h_sync_width = 4; optc1->min_v_sync_width = 1; + optc1->max_frame_count = 0xFFFFFF; + + dcn35_timing_generator_set_fgcg( + optc1, CTX->dc->debug.enable_fine_grain_clock_gating.bits.optc); } diff --git a/drivers/gpu/drm/amd/display/dc/optc/dcn42/dcn42_optc.h b/drivers/gpu/drm/amd/display/dc/optc/dcn42/dcn42_optc.h index 45d2187efaca..fc7192f01b33 100644 --- a/drivers/gpu/drm/amd/display/dc/optc/dcn42/dcn42_optc.h +++ b/drivers/gpu/drm/amd/display/dc/optc/dcn42/dcn42_optc.h @@ -119,6 +119,10 @@ SF(ODM0_OPTC_INPUT_GLOBAL_CONTROL, OPTC_UNDERFLOW_OCCURRED_STATUS, mask_sh),\ SF(ODM0_OPTC_INPUT_GLOBAL_CONTROL, OPTC_DOUBLE_BUFFER_PENDING, mask_sh),\ SF(ODM0_OPTC_INPUT_GLOBAL_CONTROL, OPTC_UNDERFLOW_CLEAR, mask_sh),\ + SF(ODM0_OPTC_RSMU_UNDERFLOW, OPTC_RSMU_UNDERFLOW_OCCURRED_STATUS, mask_sh),\ + SF(ODM0_OPTC_RSMU_UNDERFLOW, OPTC_RSMU_UNDERFLOW_CLEAR, mask_sh),\ + SF(ODM0_OPTC_RSMU_UNDERFLOW, OPTC_RSMU_UNDERFLOW_INT_EN, mask_sh),\ + SF(ODM0_OPTC_RSMU_UNDERFLOW, OPTC_RSMU_UNDERFLOW_INT_STATUS, mask_sh),\ SF(VTG0_CONTROL, VTG0_ENABLE, mask_sh),\ SF(VTG0_CONTROL, VTG0_FP2, mask_sh),\ SF(VTG0_CONTROL, VTG0_VCOUNT_INIT, mask_sh),\ @@ -202,10 +206,15 @@ SF(OTG0_OTG_PWA_FRAME_SYNC_CONTROL, OTG_PWA_FRAME_SYNC_EN, mask_sh),\ SF(OTG0_OTG_PWA_FRAME_SYNC_CONTROL, OTG_PWA_FRAME_SYNC_VCOUNT_MODE, mask_sh),\ SF(OTG0_OTG_PWA_FRAME_SYNC_CONTROL, OTG_PWA_FRAME_SYNC_LINE, mask_sh),\ - SF(OTG0_INTERRUPT_DEST, OTG0_IHC_OTG_VERTICAL_INTERRUPT2_DEST, mask_sh) + SF(OTG0_INTERRUPT_DEST, OTG0_IHC_OTG_VERTICAL_INTERRUPT2_DEST, mask_sh),\ + SF(OPTC_CLOCK_CONTROL, OPTC_FGCG_REP_DIS, mask_sh) void dcn42_timing_generator_init(struct optc *optc1); void optc42_enable_pwa(struct timing_generator *optc, struct otc_pwa_frame_sync *pwa_sync_param); void optc42_disable_pwa(struct timing_generator *optc); - +void optc42_tg_init(struct timing_generator *optc); +void optc42_clear_optc_underflow(struct timing_generator *optc); +bool optc42_is_optc_underflow_occurred(struct timing_generator *optc); +bool optc42_disable_crtc(struct timing_generator *optc); +void optc42_lock_doublebuffer_enable(struct timing_generator *optc); #endif /* __DC_OPTC_DCN42_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c index 8e41367cf238..aef187bcf5c3 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn42/dcn42_resource.c @@ -53,7 +53,7 @@ #include "dce/dce_audio.h" #include "dce/dce_hwseq.h" #include "clk_mgr.h" -#include "virtual/virtual_stream_encoder.h" +#include "dio/virtual/virtual_stream_encoder.h" #include "dml/display_mode_vba.h" #include "dcn42/dcn42_dccg.h" #include "dcn10/dcn10_resource.h" @@ -666,6 +666,7 @@ static const struct resource_caps res_cap_dcn42 = { .num_vmid = 16, .num_mpc_3dlut = 2, .num_dsc = 4, + .num_rmcm = 2, }; static const struct dc_plane_cap plane_cap = { @@ -755,6 +756,7 @@ static const struct dc_debug_options debug_defaults_drv = { .dcc_meta_propagation_delay_us = 10, .disable_timeout = true, .min_disp_clk_khz = 50000, + .static_screen_wait_frames = 2, .disable_z10 = false, .ignore_pg = true, .disable_stutter_for_wm_program = true, @@ -2302,14 +2304,12 @@ static bool dcn42_resource_construct( dc->dml2_options.max_segments_per_hubp = 24; dc->dml2_options.det_segment_size = DCN42_CRB_SEGMENT_SIZE_KB; + dc->dml2_options.gpuvm_enable = true; + dc->dml2_options.hostvm_enable = true; /* SPL */ dc->caps.scl_caps.sharpener_support = true; - /* init DC limited DML2 options */ - memcpy(&dc->dml2_dc_power_options, &dc->dml2_options, sizeof(struct dml2_configuration_options)); - dc->dml2_dc_power_options.use_clock_dc_limits = true; - return true; create_fail: diff --git a/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn401/dcn401_soc_and_ip_translator.c b/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn401/dcn401_soc_and_ip_translator.c index 3190c76eb482..1b397fa7e05c 100644 --- a/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn401/dcn401_soc_and_ip_translator.c +++ b/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn401/dcn401_soc_and_ip_translator.c @@ -102,6 +102,9 @@ static void dcn401_convert_dc_clock_table_to_soc_bb_clock_table( } } else { dml_clk_table->uclk.clk_values_khz[i] = dc_clk_table->entries[i].memclk_mhz * 1000; + #ifdef ENABLE_WCK + dml_clk_table->wck_ratio.clk_values_khz[i] = dc_clk_table->entries[i].wck_ratio; + #endif } } else { dml_clk_table->uclk.clk_values_khz[i] = 0; diff --git a/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.c b/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.c index 87ac56739030..fd9c24b5df53 100644 --- a/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.c +++ b/drivers/gpu/drm/amd/display/dc/soc_and_ip_translator/dcn42/dcn42_soc_and_ip_translator.c @@ -3,7 +3,7 @@ // Copyright 2025 Advanced Micro Devices, Inc. #include "dcn42_soc_and_ip_translator.h" -#include "soc_and_ip_translator/dcn401/dcn401_soc_and_ip_translator.h" +#include "../dcn401/dcn401_soc_and_ip_translator.h" #include "bounding_boxes/dcn42_soc_bb.h" /* soc_and_ip_translator component used to get up-to-date values for bounding box. @@ -11,10 +11,12 @@ * This component provides an interface to get DCN-specific bounding box values. */ -static void get_default_soc_bb(struct dml2_soc_bb *soc_bb) +static void get_default_soc_bb(struct dml2_soc_bb *soc_bb, const struct dc *dc) { - memcpy(soc_bb, &dml2_socbb_dcn42, sizeof(struct dml2_soc_bb)); - memcpy(&soc_bb->qos_parameters, &dml_dcn42_variant_a_soc_qos_params, sizeof(struct dml2_soc_qos_parameters)); + { + memcpy(soc_bb, &dml2_socbb_dcn42, sizeof(struct dml2_soc_bb)); + memcpy(&soc_bb->qos_parameters, &dml_dcn42_variant_a_soc_qos_params, sizeof(struct dml2_soc_qos_parameters)); + } } /* @@ -162,7 +164,7 @@ static void apply_soc_bb_updates(struct dml2_soc_bb *soc_bb, const struct dc *dc void dcn42_get_soc_bb(struct dml2_soc_bb *soc_bb, const struct dc *dc, const struct dml2_configuration_options *config) { //get default soc_bb with static values - get_default_soc_bb(soc_bb); + get_default_soc_bb(soc_bb, dc); //update soc_bb values with more accurate values apply_soc_bb_updates(soc_bb, dc, config); } diff --git a/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h b/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h index f67a0ffb9335..0b04f985b15e 100644 --- a/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h +++ b/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h @@ -909,7 +909,8 @@ union dmub_fw_meta_feature_bits { struct { uint32_t shared_state_link_detection : 1; /**< 1 supports link detection via shared state */ uint32_t cursor_offload_v1_support: 1; /**< 1 supports cursor offload */ - uint32_t reserved : 30; + uint32_t inbox0_lock_support: 1; /**< 1 supports inbox0 lock mechanism */ + uint32_t reserved : 29; } bits; /**< status bits */ uint32_t all; /**< 32-bit access to status bits */ }; @@ -1535,14 +1536,12 @@ enum dmub_gpint_command { * 1 - Enable ips measurement */ DMUB_GPINT__IPS_RESIDENCY = 121, - /** * DESC: Enable measurements for various task duration * ARGS: 0 - Disable measurement * 1 - Enable measurement */ DMUB_GPINT__TRACE_DMUB_WAKE_ACTIVITY = 123, - /** * DESC: Gets IPS residency in microseconds * ARGS: 0 - Return IPS1 residency @@ -1552,21 +1551,18 @@ enum dmub_gpint_command { * RETURN: Total residency in microseconds - lower 32 bits */ DMUB_GPINT__GET_IPS_RESIDENCY_DURATION_US_LO = 124, - /** * DESC: Gets IPS1 histogram counts * ARGS: Bucket index * RETURN: Total count for the bucket */ DMUB_GPINT__GET_IPS1_HISTOGRAM_COUNTER = 125, - /** * DESC: Gets IPS2 histogram counts * ARGS: Bucket index * RETURN: Total count for the bucket */ DMUB_GPINT__GET_IPS2_HISTOGRAM_COUNTER = 126, - /** * DESC: Gets IPS residency * ARGS: 0 - Return IPS1 residency @@ -1576,21 +1572,18 @@ enum dmub_gpint_command { * RETURN: Total residency in milli-percent. */ DMUB_GPINT__GET_IPS_RESIDENCY_PERCENT = 127, - /** * DESC: Gets IPS1_RCG histogram counts * ARGS: Bucket index * RETURN: Total count for the bucket */ DMUB_GPINT__GET_IPS1_RCG_HISTOGRAM_COUNTER = 128, - /** * DESC: Gets IPS1_ONO2_ON histogram counts * ARGS: Bucket index * RETURN: Total count for the bucket */ DMUB_GPINT__GET_IPS1_ONO2_ON_HISTOGRAM_COUNTER = 129, - /** * DESC: Gets IPS entry counter during residency measurement * ARGS: 0 - Return IPS1 entry counts @@ -1600,7 +1593,6 @@ enum dmub_gpint_command { * RETURN: Entry counter for selected IPS mode */ DMUB_GPINT__GET_IPS_RESIDENCY_ENTRY_COUNTER = 130, - /** * DESC: Gets IPS inactive residency in microseconds * ARGS: 0 - Return IPS1_MAX residency @@ -1610,7 +1602,6 @@ enum dmub_gpint_command { * RETURN: Total inactive residency in microseconds - lower 32 bits */ DMUB_GPINT__GET_IPS_INACTIVE_RESIDENCY_DURATION_US_LO = 131, - /** * DESC: Gets IPS inactive residency in microseconds * ARGS: 0 - Return IPS1_MAX residency @@ -1620,7 +1611,6 @@ enum dmub_gpint_command { * RETURN: Total inactive residency in microseconds - upper 32 bits */ DMUB_GPINT__GET_IPS_INACTIVE_RESIDENCY_DURATION_US_HI = 132, - /** * DESC: Gets IPS residency in microseconds * ARGS: 0 - Return IPS1 residency @@ -1679,7 +1669,7 @@ union dmub_inbox0_cmd_lock_hw { uint32_t lock: 1; /**< Lock */ uint32_t should_release: 1; /**< Release */ - uint32_t reserved: 7; /**< Reserved for extending more clients, HW, etc. */ + uint32_t reserved: 7; /**< Reserved for extending more clients, HW, etc. */ } bits; uint32_t all; }; @@ -2614,9 +2604,9 @@ struct dmub_fams2_stream_static_state { uint8_t allow_to_target_delay_otg_vlines; // time from allow vline to target vline union { struct { - uint8_t is_drr: 1; // stream is DRR enabled - uint8_t clamp_vtotal_min: 1; // clamp vtotal to min instead of nominal - uint8_t min_ttu_vblank_usable: 1; // if min ttu vblank is above wm, no force pstate is needed in blank + uint8_t is_drr : 1; // stream is DRR enabled + uint8_t clamp_vtotal_min : 1; // clamp vtotal to min instead of nominal + uint8_t min_ttu_vblank_usable : 1; // if min ttu vblank is above wm, no force pstate is needed in blank } bits; uint8_t all; } config; @@ -4660,6 +4650,18 @@ struct dmub_rb_cmd_replay_enable_data { * This does not support HDMI/DP2 for now. */ uint8_t phy_rate; + /** + * @hpo_stream_enc_inst: HPO stream encoder instance + */ + uint8_t hpo_stream_enc_inst; + /** + * @hpo_link_enc_inst: HPO link encoder instance + */ + uint8_t hpo_link_enc_inst; + /** + * @pad: Align structure to 4 byte boundary. + */ + uint8_t pad[2]; }; /** @@ -5273,8 +5275,8 @@ enum dmub_cmd_lsdma_type { */ DMUB_CMD__LSDMA_LINEAR_COPY = 1, /** - * LSDMA copies data from source to destination linearly in sub window - */ + * LSDMA copies data from source to destination linearly in sub window + */ DMUB_CMD__LSDMA_LINEAR_SUB_WINDOW_COPY = 2, /** * Send the tiled-to-tiled copy command diff --git a/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_2_0_offset.h b/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_2_0_offset.h index 825201f4e113..52fbf2dc1899 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_2_0_offset.h +++ b/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_2_0_offset.h @@ -9010,6 +9010,8 @@ // base address: 0x0 #define regODM0_OPTC_INPUT_GLOBAL_CONTROL 0x1aca #define regODM0_OPTC_INPUT_GLOBAL_CONTROL_BASE_IDX 2 +#define regODM0_OPTC_RSMU_UNDERFLOW 0x1acb +#define regODM0_OPTC_RSMU_UNDERFLOW_BASE_IDX 2 #define regODM0_OPTC_UNDERFLOW_THRESHOLD 0x1acc #define regODM0_OPTC_UNDERFLOW_THRESHOLD_BASE_IDX 2 #define regODM0_OPTC_DATA_SOURCE_SELECT 0x1acd diff --git a/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_2_0_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_2_0_sh_mask.h index 4ed96244f61b..01fb53093369 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_2_0_sh_mask.h +++ b/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_4_2_0_sh_mask.h @@ -33583,6 +33583,15 @@ #define ODM0_OPTC_INPUT_GLOBAL_CONTROL__OPTC_UNDERFLOW_CLEAR_MASK 0x00001000L #define ODM0_OPTC_INPUT_GLOBAL_CONTROL__OPTC_UNDERFLOW_OCCURRED_CURRENT_MASK 0x00002000L #define ODM0_OPTC_INPUT_GLOBAL_CONTROL__OPTC_DOUBLE_BUFFER_PENDING_MASK 0x80000000L +//ODM0_OPTC_RSMU_UNDERFLOW +#define ODM0_OPTC_RSMU_UNDERFLOW__OPTC_RSMU_UNDERFLOW_INT_EN__SHIFT 0x0 +#define ODM0_OPTC_RSMU_UNDERFLOW__OPTC_RSMU_UNDERFLOW_OCCURRED_STATUS__SHIFT 0x1 +#define ODM0_OPTC_RSMU_UNDERFLOW__OPTC_RSMU_UNDERFLOW_INT_STATUS__SHIFT 0x2 +#define ODM0_OPTC_RSMU_UNDERFLOW__OPTC_RSMU_UNDERFLOW_CLEAR__SHIFT 0x3 +#define ODM0_OPTC_RSMU_UNDERFLOW__OPTC_RSMU_UNDERFLOW_INT_EN_MASK 0x00000001L +#define ODM0_OPTC_RSMU_UNDERFLOW__OPTC_RSMU_UNDERFLOW_OCCURRED_STATUS_MASK 0x00000002L +#define ODM0_OPTC_RSMU_UNDERFLOW__OPTC_RSMU_UNDERFLOW_INT_STATUS_MASK 0x00000004L +#define ODM0_OPTC_RSMU_UNDERFLOW__OPTC_RSMU_UNDERFLOW_CLEAR_MASK 0x00000008L //ODM0_OPTC_UNDERFLOW_THRESHOLD #define ODM0_OPTC_UNDERFLOW_THRESHOLD__OPTC_UNDERFLOW_THRESHOLD__SHIFT 0x0 #define ODM0_OPTC_UNDERFLOW_THRESHOLD__OPTC_UNDERFLOW_THRESHOLD_MASK 0x01FFFFFFL -- cgit v1.2.3 From 9e9fbebde4a42268a39e4d9355fd038313f5da64 Mon Sep 17 00:00:00 2001 From: Taimur Hassan Date: Fri, 27 Feb 2026 16:35:12 -0500 Subject: drm/amd/display: [FW Promotion] Release 0.1.50.0 [WHAT] Introduce DMUB BOOT_TIME_CRC_INIT command. Acked-by: Aurabindo Pillai Signed-off-by: Taimur Hassan Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h | 32 +++++++++++++++++++++++++ 1 file changed, 32 insertions(+) diff --git a/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h b/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h index 0b04f985b15e..e11e32afac6b 100644 --- a/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h +++ b/drivers/gpu/drm/amd/display/dmub/inc/dmub_cmd.h @@ -1892,6 +1892,11 @@ enum dmub_cmd_type { */ DMUB_CMD__IHC = 95, + /** + * Command type use for boot time crc commands. + */ + DMUB_CMD__BOOT_TIME_CRC = 96, + /** * Command type use for VBIOS shared commands. */ @@ -6838,6 +6843,29 @@ struct dmub_rb_cmd_pr_general_cmd { struct dmub_cmd_pr_general_cmd_data data; }; +/** + * Command type of a DMUB_CMD__BOOT_TIME_CRC command + */ +enum dmub_cmd_boot_time_crc_type { + DMUB_CMD__BOOT_TIME_CRC_INIT_MEM = 0 +}; + +/** + * Data passed from driver to FW in a DMUB_CMD__BOOT_TIME_CRC_INIT command. + */ +struct dmub_cmd_boot_time_crc_init_data { + union dmub_addr buffer_addr; + uint32_t buffer_size; +}; + +/** + * Definition of a DMUB_CMD__BOOT_TIME_CRC_INIT command. + */ +struct dmub_rb_cmd_boot_time_crc_init { + struct dmub_cmd_header header; + struct dmub_cmd_boot_time_crc_init_data data; +}; + /** * union dmub_rb_cmd - DMUB inbox command. */ @@ -7199,6 +7227,10 @@ union dmub_rb_cmd { * Definition of a DMUB_CMD__IHC command. */ struct dmub_rb_cmd_ihc ihc; + /** + * Definition of a DMUB_CMD__BOOT_TIME_CRC_INIT command. + */ + struct dmub_rb_cmd_boot_time_crc_init boot_time_crc_init; }; /** -- cgit v1.2.3 From 78aad33088dd7fc3e222d387a272b2e50dec3fb9 Mon Sep 17 00:00:00 2001 From: Taimur Hassan Date: Fri, 27 Feb 2026 17:40:05 -0500 Subject: drm/amd/display: Promote DC to 3.2.373 This version brings along the following updates: - [FW Promotion] Release 0.1.50.0 - Sync DCN42 with DC 3.2.373 - Add DML support for dcn42 - Enable dcn42 DC clk_mgr - Clean up unused code - Add back missing memory type in array - Fix compile warnings in dml2_0 - Check for S0i3 to be done before DCCG init on DCN21 - Add documentation and cleanup DMUB HW lock manager - Add new types to replay config - Fix HWSS v3 fast path determination - Add missing DCCG register entries for DCN20-DCN316 - Add ESD detection for replay recovery - Update underflow detection - Add COLOR_ENCODING/COLOR_RANGE to overlay planes - Add NV12/P010 formats to primary plane - Set chroma taps to 1 if luma taps are 1 - Add min clock init for DML21 mode programming - Return early from vesa replay enable function - Clean up NULL pointer warnings in dml2 Acked-by: Aurabindo Pillai Signed-off-by: Taimur Hassan Signed-off-by: Alex Hung Tested-by: Dan Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dc.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h index 8c49f8083475..c7a09724f569 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -63,7 +63,7 @@ struct dcn_dsc_reg_state; struct dcn_optc_reg_state; struct dcn_dccg_reg_state; -#define DC_VER "3.2.372" +#define DC_VER "3.2.373" /** * MAX_SURFACES - representative of the upper bound of surfaces that can be piped to a single CRTC -- cgit v1.2.3 From c24afed7de9ecce341825d8ab55a43a254348b33 Mon Sep 17 00:00:00 2001 From: Philip Yang Date: Tue, 9 Dec 2025 15:13:23 -0500 Subject: drm/amdkfd: Unreserve bo if queue update failed Error handling path should unreserve bo then return failed. Fixes: 305cd109b761 ("drm/amdkfd: Validate user queue update") Signed-off-by: Philip Yang Reviewed-by: Alex Sierra Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c index 586d409ebe4e..cc2621ae12f9 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c @@ -593,6 +593,7 @@ int pqm_update_queue_properties(struct process_queue_manager *pqm, p->queue_size)) { pr_debug("ring buf 0x%llx size 0x%llx not mapped on GPU\n", p->queue_address, p->queue_size); + amdgpu_bo_unreserve(vm->root.bo); return -EFAULT; } -- cgit v1.2.3 From d0c647a6aae2ce577372594a176c82733e1ed267 Mon Sep 17 00:00:00 2001 From: Likun Gao Date: Fri, 7 Nov 2025 11:59:05 +0800 Subject: drm/amdgpu/discovery: support new discovery binary header Support for new IP discovery binary header version 2. Signed-off-by: Likun Gao Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 128 ++++++++++++++++++-------- drivers/gpu/drm/amd/include/discovery.h | 13 +++ 2 files changed, 104 insertions(+), 37 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index c4ae1ce52a0f..163b40c8afb0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -431,14 +431,12 @@ static void amdgpu_discovery_harvest_config_quirk(struct amdgpu_device *adev) } static int amdgpu_discovery_verify_npsinfo(struct amdgpu_device *adev, - struct binary_header *bhdr) + struct table_info *info) { uint8_t *discovery_bin = adev->discovery.bin; - struct table_info *info; uint16_t checksum; uint16_t offset; - info = &bhdr->table_list[NPS_INFO]; offset = le16_to_cpu(info->offset); checksum = le16_to_cpu(info->checksum); @@ -491,6 +489,30 @@ static const char *amdgpu_discovery_get_fw_name(struct amdgpu_device *adev) } } +static int amdgpu_discovery_get_table_info(struct amdgpu_device *adev, + struct table_info **info, + uint16_t table_id) +{ + struct binary_header *bhdr = + (struct binary_header *)adev->discovery.bin; + struct binary_header_v2 *bhdrv2; + + switch (bhdr->version_major) { + case 2: + bhdrv2 = (struct binary_header_v2 *)adev->discovery.bin; + *info = &bhdrv2->table_list[table_id]; + break; + case 1: + *info = &bhdr->table_list[table_id]; + break; + default: + dev_err(adev->dev, "Invalid ip discovery table version\n"); + return -EINVAL; + } + + return 0; +} + static int amdgpu_discovery_init(struct amdgpu_device *adev) { struct table_info *info; @@ -547,7 +569,9 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev) goto out; } - info = &bhdr->table_list[IP_DISCOVERY]; + r = amdgpu_discovery_get_table_info(adev, &info, IP_DISCOVERY); + if (r) + goto out; offset = le16_to_cpu(info->offset); checksum = le16_to_cpu(info->checksum); @@ -569,7 +593,9 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev) } } - info = &bhdr->table_list[GC]; + r = amdgpu_discovery_get_table_info(adev, &info, GC); + if (r) + goto out; offset = le16_to_cpu(info->offset); checksum = le16_to_cpu(info->checksum); @@ -592,7 +618,9 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev) } } - info = &bhdr->table_list[HARVEST_INFO]; + r = amdgpu_discovery_get_table_info(adev, &info, HARVEST_INFO); + if (r) + goto out; offset = le16_to_cpu(info->offset); checksum = le16_to_cpu(info->checksum); @@ -615,7 +643,9 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev) } } - info = &bhdr->table_list[VCN_INFO]; + r = amdgpu_discovery_get_table_info(adev, &info, VCN_INFO); + if (r) + goto out; offset = le16_to_cpu(info->offset); checksum = le16_to_cpu(info->checksum); @@ -638,7 +668,9 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev) } } - info = &bhdr->table_list[MALL_INFO]; + r = amdgpu_discovery_get_table_info(adev, &info, MALL_INFO); + if (r) + goto out; offset = le16_to_cpu(info->offset); checksum = le16_to_cpu(info->checksum); @@ -770,14 +802,15 @@ static void amdgpu_discovery_read_from_harvest_table(struct amdgpu_device *adev, uint32_t *umc_harvest_count) { uint8_t *discovery_bin = adev->discovery.bin; - struct binary_header *bhdr; + struct table_info *info; struct harvest_table *harvest_info; u16 offset; int i; uint32_t umc_harvest_config = 0; - bhdr = (struct binary_header *)discovery_bin; - offset = le16_to_cpu(bhdr->table_list[HARVEST_INFO].offset); + if (amdgpu_discovery_get_table_info(adev, &info, HARVEST_INFO)) + return; + offset = le16_to_cpu(info->offset); if (!offset) { dev_err(adev->dev, "invalid harvest table offset\n"); @@ -1224,7 +1257,7 @@ static int amdgpu_discovery_sysfs_recurse(struct amdgpu_device *adev) { struct ip_discovery_top *ip_top = adev->discovery.ip_top; uint8_t *discovery_bin = adev->discovery.bin; - struct binary_header *bhdr; + struct table_info *info; struct ip_discovery_header *ihdr; struct die_header *dhdr; struct kset *die_kset = &ip_top->die_kset; @@ -1232,10 +1265,12 @@ static int amdgpu_discovery_sysfs_recurse(struct amdgpu_device *adev) size_t ip_offset; int ii, res; - bhdr = (struct binary_header *)discovery_bin; + res = amdgpu_discovery_get_table_info(adev, &info, IP_DISCOVERY); + if (res) + return res; ihdr = (struct ip_discovery_header *)(discovery_bin + - le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset)); + le16_to_cpu(info->offset)); num_dies = le16_to_cpu(ihdr->num_dies); DRM_DEBUG("number of dies: %d\n", num_dies); @@ -1426,7 +1461,7 @@ void amdgpu_discovery_dump(struct amdgpu_device *adev, struct drm_printer *p) static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev) { uint8_t num_base_address, subrev, variant; - struct binary_header *bhdr; + struct table_info *info; struct ip_discovery_header *ihdr; struct die_header *dhdr; uint8_t *discovery_bin; @@ -1451,10 +1486,12 @@ static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev) adev->sdma.sdma_mask = 0; adev->vcn.inst_mask = 0; adev->jpeg.inst_mask = 0; - bhdr = (struct binary_header *)discovery_bin; + r = amdgpu_discovery_get_table_info(adev, &info, IP_DISCOVERY); + if (r) + return r; ihdr = (struct ip_discovery_header *)(discovery_bin + - le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset)); + le16_to_cpu(info->offset)); num_dies = le16_to_cpu(ihdr->num_dies); DRM_DEBUG("number of dies: %d\n", num_dies); @@ -1627,14 +1664,15 @@ static void amdgpu_discovery_harvest_ip(struct amdgpu_device *adev) { uint8_t *discovery_bin = adev->discovery.bin; struct ip_discovery_header *ihdr; - struct binary_header *bhdr; + struct table_info *info; int vcn_harvest_count = 0; int umc_harvest_count = 0; - uint16_t offset, ihdr_ver; + uint16_t ihdr_ver; - bhdr = (struct binary_header *)discovery_bin; - offset = le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset); - ihdr = (struct ip_discovery_header *)(discovery_bin + offset); + if (amdgpu_discovery_get_table_info(adev, &info, IP_DISCOVERY)) + return; + ihdr = (struct ip_discovery_header *)(discovery_bin + + le16_to_cpu(info->offset)); ihdr_ver = le16_to_cpu(ihdr->version); /* * Harvest table does not fit Navi1x and legacy GPUs, @@ -1682,7 +1720,7 @@ union gc_info { static int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev) { uint8_t *discovery_bin = adev->discovery.bin; - struct binary_header *bhdr; + struct table_info *info; union gc_info *gc_info; u16 offset; @@ -1691,8 +1729,9 @@ static int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev) return -EINVAL; } - bhdr = (struct binary_header *)discovery_bin; - offset = le16_to_cpu(bhdr->table_list[GC].offset); + if (amdgpu_discovery_get_table_info(adev, &info, GC)) + return -EINVAL; + offset = le16_to_cpu(info->offset); if (!offset) return 0; @@ -1791,7 +1830,7 @@ union mall_info { static int amdgpu_discovery_get_mall_info(struct amdgpu_device *adev) { uint8_t *discovery_bin = adev->discovery.bin; - struct binary_header *bhdr; + struct table_info *info; union mall_info *mall_info; u32 u, mall_size_per_umc, m_s_present, half_use; u64 mall_size; @@ -1802,8 +1841,9 @@ static int amdgpu_discovery_get_mall_info(struct amdgpu_device *adev) return -EINVAL; } - bhdr = (struct binary_header *)discovery_bin; - offset = le16_to_cpu(bhdr->table_list[MALL_INFO].offset); + if (amdgpu_discovery_get_table_info(adev, &info, MALL_INFO)) + return -EINVAL; + offset = le16_to_cpu(info->offset); if (!offset) return 0; @@ -1848,7 +1888,7 @@ union vcn_info { static int amdgpu_discovery_get_vcn_info(struct amdgpu_device *adev) { uint8_t *discovery_bin = adev->discovery.bin; - struct binary_header *bhdr; + struct table_info *info; union vcn_info *vcn_info; u16 offset; int v; @@ -1868,8 +1908,9 @@ static int amdgpu_discovery_get_vcn_info(struct amdgpu_device *adev) return -EINVAL; } - bhdr = (struct binary_header *)discovery_bin; - offset = le16_to_cpu(bhdr->table_list[VCN_INFO].offset); + if (amdgpu_discovery_get_table_info(adev, &info, VCN_INFO)) + return -EINVAL; + offset = le16_to_cpu(info->offset); if (!offset) return 0; @@ -1906,14 +1947,26 @@ static int amdgpu_discovery_refresh_nps_info(struct amdgpu_device *adev, uint64_t vram_size, pos, offset; struct nps_info_header *nhdr; struct binary_header bhdr; + struct binary_header_v2 bhdrv2; uint16_t checksum; vram_size = (uint64_t)RREG32(mmRCC_CONFIG_MEMSIZE) << 20; pos = vram_size - DISCOVERY_TMR_OFFSET; amdgpu_device_vram_access(adev, pos, &bhdr, sizeof(bhdr), false); - offset = le16_to_cpu(bhdr.table_list[NPS_INFO].offset); - checksum = le16_to_cpu(bhdr.table_list[NPS_INFO].checksum); + switch (bhdr.version_major) { + case 2: + amdgpu_device_vram_access(adev, pos, &bhdrv2, sizeof(bhdrv2), false); + offset = le16_to_cpu(bhdrv2.table_list[NPS_INFO].offset); + checksum = le16_to_cpu(bhdrv2.table_list[NPS_INFO].checksum); + break; + case 1: + offset = le16_to_cpu(bhdr.table_list[NPS_INFO].offset); + checksum = le16_to_cpu(bhdr.table_list[NPS_INFO].checksum); + break; + default: + return -EINVAL; + } amdgpu_device_vram_access(adev, (pos + offset), nps_data, sizeof(*nps_data), false); @@ -1936,7 +1989,7 @@ int amdgpu_discovery_get_nps_info(struct amdgpu_device *adev, { uint8_t *discovery_bin = adev->discovery.bin; struct amdgpu_gmc_memrange *mem_ranges; - struct binary_header *bhdr; + struct table_info *info; union nps_info *nps_info; union nps_info nps_data; u16 offset; @@ -1957,14 +2010,15 @@ int amdgpu_discovery_get_nps_info(struct amdgpu_device *adev, return -EINVAL; } - bhdr = (struct binary_header *)discovery_bin; - offset = le16_to_cpu(bhdr->table_list[NPS_INFO].offset); + if (amdgpu_discovery_get_table_info(adev, &info, NPS_INFO)) + return -EINVAL; + offset = le16_to_cpu(info->offset); if (!offset) return -ENOENT; /* If verification fails, return as if NPS table doesn't exist */ - if (amdgpu_discovery_verify_npsinfo(adev, bhdr)) + if (amdgpu_discovery_verify_npsinfo(adev, info)) return -ENOENT; nps_info = (union nps_info *)(discovery_bin + offset); diff --git a/drivers/gpu/drm/amd/include/discovery.h b/drivers/gpu/drm/amd/include/discovery.h index 710e328fad48..76c9f951bc1c 100644 --- a/drivers/gpu/drm/amd/include/discovery.h +++ b/drivers/gpu/drm/amd/include/discovery.h @@ -64,6 +64,19 @@ typedef struct binary_header table_info table_list[TOTAL_TABLES]; } binary_header; +typedef struct binary_header_v2 +{ + /* psp structure should go at the top of this structure */ + uint32_t binary_signature; /* 0x7, 0x14, 0x21, 0x28 */ + uint16_t version_major; /* 0x02 */ + uint16_t version_minor; + uint16_t binary_checksum; /* Byte sum of the binary after this field */ + uint16_t binary_size; /* Binary Size*/ + uint16_t num_tables; + uint16_t padding; + table_info table_list[] __counted_by(num_tables); +} binary_header_v2; + typedef struct die_info { uint16_t die_id; -- cgit v1.2.3 From 47ab777c16c71ba033325593cd833bded65f5943 Mon Sep 17 00:00:00 2001 From: Likun Gao Date: Thu, 5 Mar 2026 14:34:56 -0500 Subject: drm/amdgpu/discovery: use common function to check discovery table Use an common function to check the validation of discovery table. Signed-off-by: Likun Gao Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 206 +++++++++++--------------- 1 file changed, 84 insertions(+), 122 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index 163b40c8afb0..6dd2886e6e7c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -513,15 +513,95 @@ static int amdgpu_discovery_get_table_info(struct amdgpu_device *adev, return 0; } -static int amdgpu_discovery_init(struct amdgpu_device *adev) +static int amdgpu_discovery_table_check(struct amdgpu_device *adev, + uint8_t *discovery_bin, + uint16_t table_id) { + int r, act_val, exp_val, table_size; + uint16_t offset, checksum; struct table_info *info; + bool check_table = true; + char *table_name; + + r = amdgpu_discovery_get_table_info(adev, &info, table_id); + if (r) + return r; + offset = le16_to_cpu(info->offset); + checksum = le16_to_cpu(info->checksum); + + switch (table_id) { + case IP_DISCOVERY: + struct ip_discovery_header *ihdr = + (struct ip_discovery_header *)(discovery_bin + offset); + act_val = le32_to_cpu(ihdr->signature); + exp_val = DISCOVERY_TABLE_SIGNATURE; + table_size = le16_to_cpu(ihdr->size); + table_name = "data table"; + break; + case GC: + struct gpu_info_header *ghdr = + (struct gpu_info_header *)(discovery_bin + offset); + act_val = le32_to_cpu(ghdr->table_id); + exp_val = GC_TABLE_ID; + table_size = le16_to_cpu(ghdr->size); + table_name = "gc table"; + break; + case HARVEST_INFO: + struct harvest_info_header *hhdr = + (struct harvest_info_header *)(discovery_bin + offset); + act_val = le32_to_cpu(hhdr->signature); + exp_val = HARVEST_TABLE_SIGNATURE; + table_size = sizeof(struct harvest_table); + table_name = "harvest table"; + break; + case VCN_INFO: + struct vcn_info_header *vhdr = + (struct vcn_info_header *)(discovery_bin + offset); + act_val = le32_to_cpu(vhdr->table_id); + exp_val = VCN_INFO_TABLE_ID; + table_size = le32_to_cpu(vhdr->size_bytes); + table_name = "vcn table"; + break; + case MALL_INFO: + struct mall_info_header *mhdr = + (struct mall_info_header *)(discovery_bin + offset); + act_val = le32_to_cpu(mhdr->table_id); + exp_val = MALL_INFO_TABLE_ID; + table_size = le32_to_cpu(mhdr->size_bytes); + table_name = "mall table"; + check_table = false; + break; + default: + dev_err(adev->dev, "invalid ip discovery table id %d specified\n", table_id); + check_table = false; + break; + } + + if (check_table && offset) { + if (act_val != exp_val) { + dev_err(adev->dev, "invalid ip discovery %s signature\n", table_name); + return -EINVAL; + } + + if (!amdgpu_discovery_verify_checksum(adev, discovery_bin + offset, + table_size, checksum)) { + dev_err(adev->dev, "invalid ip discovery %s checksum\n", table_name); + return -EINVAL; + } + } + + return 0; +} + +static int amdgpu_discovery_init(struct amdgpu_device *adev) +{ struct binary_header *bhdr; uint8_t *discovery_bin; const char *fw_name; uint16_t offset; uint16_t size; uint16_t checksum; + uint16_t table_id; int r; adev->discovery.bin = kzalloc(DISCOVERY_TMR_SIZE, GFP_KERNEL); @@ -569,128 +649,10 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev) goto out; } - r = amdgpu_discovery_get_table_info(adev, &info, IP_DISCOVERY); - if (r) - goto out; - offset = le16_to_cpu(info->offset); - checksum = le16_to_cpu(info->checksum); - - if (offset) { - struct ip_discovery_header *ihdr = - (struct ip_discovery_header *)(discovery_bin + offset); - if (le32_to_cpu(ihdr->signature) != DISCOVERY_TABLE_SIGNATURE) { - dev_err(adev->dev, "invalid ip discovery data table signature\n"); - r = -EINVAL; - goto out; - } - - if (!amdgpu_discovery_verify_checksum(adev, discovery_bin + offset, - le16_to_cpu(ihdr->size), - checksum)) { - dev_err(adev->dev, "invalid ip discovery data table checksum\n"); - r = -EINVAL; - goto out; - } - } - - r = amdgpu_discovery_get_table_info(adev, &info, GC); - if (r) - goto out; - offset = le16_to_cpu(info->offset); - checksum = le16_to_cpu(info->checksum); - - if (offset) { - struct gpu_info_header *ghdr = - (struct gpu_info_header *)(discovery_bin + offset); - - if (le32_to_cpu(ghdr->table_id) != GC_TABLE_ID) { - dev_err(adev->dev, "invalid ip discovery gc table id\n"); - r = -EINVAL; - goto out; - } - - if (!amdgpu_discovery_verify_checksum(adev, discovery_bin + offset, - le32_to_cpu(ghdr->size), - checksum)) { - dev_err(adev->dev, "invalid gc data table checksum\n"); - r = -EINVAL; - goto out; - } - } - - r = amdgpu_discovery_get_table_info(adev, &info, HARVEST_INFO); - if (r) - goto out; - offset = le16_to_cpu(info->offset); - checksum = le16_to_cpu(info->checksum); - - if (offset) { - struct harvest_info_header *hhdr = - (struct harvest_info_header *)(discovery_bin + offset); - - if (le32_to_cpu(hhdr->signature) != HARVEST_TABLE_SIGNATURE) { - dev_err(adev->dev, "invalid ip discovery harvest table signature\n"); - r = -EINVAL; - goto out; - } - - if (!amdgpu_discovery_verify_checksum(adev, - discovery_bin + offset, - sizeof(struct harvest_table), checksum)) { - dev_err(adev->dev, "invalid harvest data table checksum\n"); - r = -EINVAL; - goto out; - } - } - - r = amdgpu_discovery_get_table_info(adev, &info, VCN_INFO); - if (r) - goto out; - offset = le16_to_cpu(info->offset); - checksum = le16_to_cpu(info->checksum); - - if (offset) { - struct vcn_info_header *vhdr = - (struct vcn_info_header *)(discovery_bin + offset); - - if (le32_to_cpu(vhdr->table_id) != VCN_INFO_TABLE_ID) { - dev_err(adev->dev, "invalid ip discovery vcn table id\n"); - r = -EINVAL; - goto out; - } - - if (!amdgpu_discovery_verify_checksum(adev, - discovery_bin + offset, - le32_to_cpu(vhdr->size_bytes), checksum)) { - dev_err(adev->dev, "invalid vcn data table checksum\n"); - r = -EINVAL; - goto out; - } - } - - r = amdgpu_discovery_get_table_info(adev, &info, MALL_INFO); - if (r) - goto out; - offset = le16_to_cpu(info->offset); - checksum = le16_to_cpu(info->checksum); - - if (0 && offset) { - struct mall_info_header *mhdr = - (struct mall_info_header *)(discovery_bin + offset); - - if (le32_to_cpu(mhdr->table_id) != MALL_INFO_TABLE_ID) { - dev_err(adev->dev, "invalid ip discovery mall table id\n"); - r = -EINVAL; - goto out; - } - - if (!amdgpu_discovery_verify_checksum(adev, - discovery_bin + offset, - le32_to_cpu(mhdr->size_bytes), checksum)) { - dev_err(adev->dev, "invalid mall data table checksum\n"); - r = -EINVAL; + for (table_id = 0; table_id <= MALL_INFO; table_id++) { + r = amdgpu_discovery_table_check(adev, discovery_bin, table_id); + if (r) goto out; - } } return 0; -- cgit v1.2.3 From 01bdc7e219c4e10166a2eb4642e9e55af33b8ff5 Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Tue, 10 Feb 2026 04:08:54 +0800 Subject: drm/amdgpu: New interface to get IP discovery binary v3 Implement a driver path to read the IP discovery binary offset and size from DRIVER_SCRATCH registers BIOS signals usage by setting a feature flag that instructs the driver to use this method. Otherwise, fallback to legacy approach. v2: Simplify discovery offset/size retrieval in get_tmr_info v3: Update get_tmr_info to cover discovery offset and size retrieval for both bare-metal and sriov Signed-off-by: Hawking Zhang Reviewed-by: Likun Gao Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 165 ++++++++++++++------------ drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h | 1 + 2 files changed, 93 insertions(+), 73 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index 6dd2886e6e7c..0ea3c8ad158c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -132,6 +132,7 @@ MODULE_FIRMWARE("amdgpu/picasso_ip_discovery.bin"); MODULE_FIRMWARE("amdgpu/arcturus_ip_discovery.bin"); MODULE_FIRMWARE("amdgpu/aldebaran_ip_discovery.bin"); +/* Note: These registers are consistent across all the SOCs */ #define mmIP_DISCOVERY_VERSION 0x16A00 #define mmRCC_CONFIG_MEMSIZE 0xde3 #define mmMP0_SMN_C2PMSG_33 0x16061 @@ -139,6 +140,10 @@ MODULE_FIRMWARE("amdgpu/aldebaran_ip_discovery.bin"); #define mmMM_INDEX_HI 0x6 #define mmMM_DATA 0x1 +#define mmDRIVER_SCRATCH_0 0x94 +#define mmDRIVER_SCRATCH_1 0x95 +#define mmDRIVER_SCRATCH_2 0x96 + static const char *hw_id_names[HW_ID_MAX] = { [MP1_HWID] = "MP1", [MP2_HWID] = "MP2", @@ -253,39 +258,12 @@ static int hw_id_map[MAX_HWIP] = { [ATU_HWIP] = ATU_HWID, }; -static int amdgpu_discovery_read_binary_from_sysmem(struct amdgpu_device *adev, uint8_t *binary) +static int amdgpu_discovery_get_tmr_info(struct amdgpu_device *adev, + bool *is_tmr_in_sysmem) { - u64 tmr_offset, tmr_size, pos; - void *discv_regn; - int ret; - - ret = amdgpu_acpi_get_tmr_info(adev, &tmr_offset, &tmr_size); - if (ret) - return ret; - - pos = tmr_offset + tmr_size - DISCOVERY_TMR_OFFSET; - - /* This region is read-only and reserved from system use */ - discv_regn = memremap(pos, adev->discovery.size, MEMREMAP_WC); - if (discv_regn) { - memcpy(binary, discv_regn, adev->discovery.size); - memunmap(discv_regn); - return 0; - } - - return -ENOENT; -} - -#define IP_DISCOVERY_V2 2 -#define IP_DISCOVERY_V4 4 - -static int amdgpu_discovery_read_binary_from_mem(struct amdgpu_device *adev, - uint8_t *binary) -{ - bool sz_valid = true; - uint64_t vram_size; - int i, ret = 0; - u32 msg; + u64 vram_size, tmr_offset, tmr_size; + u32 msg, tmr_offset_lo, tmr_offset_hi; + int i, ret; if (!amdgpu_sriov_vf(adev)) { /* It can take up to two second for IFWI init to complete on some dGPUs, @@ -305,51 +283,93 @@ static int amdgpu_discovery_read_binary_from_mem(struct amdgpu_device *adev, } vram_size = RREG32(mmRCC_CONFIG_MEMSIZE); - if (!vram_size || vram_size == U32_MAX) - sz_valid = false; + if (vram_size == U32_MAX) + return -ENXIO; + else if (!vram_size) + *is_tmr_in_sysmem = true; else - vram_size <<= 20; - - /* - * If in VRAM, discovery TMR is marked for reservation. If it is in system mem, - * then it is not required to be reserved. - */ - if (sz_valid) { - if (amdgpu_sriov_vf(adev) && adev->virt.is_dynamic_crit_regn_enabled) { - /* For SRIOV VFs with dynamic critical region enabled, - * we will get the IPD binary via below call. - * If dynamic critical is disabled, fall through to normal seq. - */ - if (amdgpu_virt_get_dynamic_data_info(adev, - AMD_SRIOV_MSG_IPD_TABLE_ID, binary, - &adev->discovery.size)) { - dev_err(adev->dev, - "failed to read discovery info from dynamic critical region."); - ret = -EINVAL; - goto exit; - } - } else { - uint64_t pos = vram_size - DISCOVERY_TMR_OFFSET; + *is_tmr_in_sysmem = false; - amdgpu_device_vram_access(adev, pos, (uint32_t *)binary, - adev->discovery.size, false); - adev->discovery.reserve_tmr = true; + /* init the default tmr size and offset */ + adev->discovery.size = DISCOVERY_TMR_SIZE; + if (vram_size) + adev->discovery.offset = (vram_size << 20) - DISCOVERY_TMR_OFFSET; + + if (amdgpu_sriov_vf(adev) && adev->virt.is_dynamic_crit_regn_enabled) { + adev->discovery.offset = + adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_IPD_TABLE_ID].offset; + adev->discovery.size = + adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_IPD_TABLE_ID].size_kb << 10; + if (!adev->discovery.offset || !adev->discovery.size) + return -EINVAL; + } else { + tmr_size = RREG32(mmDRIVER_SCRATCH_2); + if (tmr_size) { + /* It's preferred to transition to PSP mailbox reg interface + * for both bare-metal and passthrough if available */ + adev->discovery.size = (u32)tmr_size; + tmr_offset_lo = RREG32(mmDRIVER_SCRATCH_0); + tmr_offset_hi = RREG32(mmDRIVER_SCRATCH_1); + adev->discovery.offset = ((u64)le32_to_cpu(tmr_offset_hi) << 32 | + le32_to_cpu(tmr_offset_lo)); + } else if (!vram_size) { + /* fall back to apci approach to query tmr offset if vram_size is 0 */ + ret = amdgpu_acpi_get_tmr_info(adev, &tmr_offset, &tmr_size); + if (ret) + return ret; + adev->discovery.size = (u32)tmr_size; + adev->discovery.offset = tmr_offset + tmr_size - DISCOVERY_TMR_OFFSET; } + } + + adev->discovery.bin = kzalloc(adev->discovery.size, GFP_KERNEL); + if (!adev->discovery.bin) + return -ENOMEM; + adev->discovery.debugfs_blob.data = adev->discovery.bin; + adev->discovery.debugfs_blob.size = adev->discovery.size; + + return 0; +} + +static int amdgpu_discovery_read_binary_from_sysmem(struct amdgpu_device *adev, uint8_t *binary) +{ + void *discv_regn; + + /* This region is read-only and reserved from system use */ + discv_regn = memremap(adev->discovery.offset, adev->discovery.size, MEMREMAP_WC); + if (discv_regn) { + memcpy(binary, discv_regn, adev->discovery.size); + memunmap(discv_regn); + return 0; + } + + return -ENOENT; +} + +#define IP_DISCOVERY_V2 2 +#define IP_DISCOVERY_V4 4 + +static int amdgpu_discovery_read_binary_from_mem(struct amdgpu_device *adev, + uint8_t *binary, + bool is_tmr_in_sysmem) +{ + int ret = 0; + + if (!is_tmr_in_sysmem) { + amdgpu_device_vram_access(adev, adev->discovery.offset, + (uint32_t *)binary, + adev->discovery.size, false); + adev->discovery.reserve_tmr = true; } else { ret = amdgpu_discovery_read_binary_from_sysmem(adev, binary); } - if (ret) - dev_err(adev->dev, - "failed to read discovery info from memory, vram size read: %llx", - vram_size); -exit: return ret; } static int amdgpu_discovery_read_binary_from_file(struct amdgpu_device *adev, - uint8_t *binary, - const char *fw_name) + uint8_t *binary, + const char *fw_name) { const struct firmware *fw; int r; @@ -602,14 +622,12 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev) uint16_t size; uint16_t checksum; uint16_t table_id; + bool is_tmr_in_sysmem; int r; - adev->discovery.bin = kzalloc(DISCOVERY_TMR_SIZE, GFP_KERNEL); - if (!adev->discovery.bin) - return -ENOMEM; - adev->discovery.size = DISCOVERY_TMR_SIZE; - adev->discovery.debugfs_blob.data = adev->discovery.bin; - adev->discovery.debugfs_blob.size = adev->discovery.size; + r = amdgpu_discovery_get_tmr_info(adev, &is_tmr_in_sysmem); + if (r) + return r; discovery_bin = adev->discovery.bin; /* Read from file if it is the preferred option */ @@ -622,7 +640,8 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev) goto out; } else { drm_dbg(&adev->ddev, "use ip discovery information from memory"); - r = amdgpu_discovery_read_binary_from_mem(adev, discovery_bin); + r = amdgpu_discovery_read_binary_from_mem(adev, discovery_bin, + is_tmr_in_sysmem); if (r) goto out; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h index c8242992c912..a7aeb47887a3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h @@ -35,6 +35,7 @@ struct drm_printer; struct amdgpu_discovery_info { struct debugfs_blob_wrapper debugfs_blob; struct ip_discovery_top *ip_top; + uint64_t offset; uint32_t size; uint8_t *bin; bool reserve_tmr; -- cgit v1.2.3 From c81761f2334f479ac865ab631e5166c15227adde Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Mon, 2 Feb 2026 03:32:52 +0800 Subject: drm/amdgpu: Use memcpy to update IPD table for sriov guest On some hardware configuration, sriov guests cannot access mm_index and mm_data. Update the IPD table via memcpy in these cases Signed-off-by: Hawking Zhang Reviewed-by: Likun Gao Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 13 +++++++++---- drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h | 4 ++++ drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h | 3 ++- 3 files changed, 15 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index 0ea3c8ad158c..9b0a3107bc90 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -356,10 +356,15 @@ static int amdgpu_discovery_read_binary_from_mem(struct amdgpu_device *adev, int ret = 0; if (!is_tmr_in_sysmem) { - amdgpu_device_vram_access(adev, adev->discovery.offset, - (uint32_t *)binary, - adev->discovery.size, false); - adev->discovery.reserve_tmr = true; + if (amdgpu_sriov_vf(adev) && + amdgpu_sriov_xgmi_connected_to_cpu(adev)) { + ret = amdgpu_discovery_read_binary_from_sysmem(adev, binary); + } else { + amdgpu_device_vram_access(adev, adev->discovery.offset, + (uint32_t *)binary, + adev->discovery.size, false); + adev->discovery.reserve_tmr = true; + } } else { ret = amdgpu_discovery_read_binary_from_sysmem(adev, binary); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h index 886fbce0bfd1..9da0c6e9b869 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h @@ -162,6 +162,7 @@ enum AMDGIM_FEATURE_FLAG { AMDGIM_FEATURE_RAS_TELEMETRY = (1 << 10), AMDGIM_FEATURE_RAS_CPER = (1 << 11), AMDGIM_FEATURE_XGMI_TA_EXT_PEER_LINK = (1 << 12), + AMDGIM_FEATURE_XGMI_CONNECTED_TO_CPU = (1 << 13), }; enum AMDGIM_REG_ACCESS_FLAG { @@ -412,6 +413,9 @@ struct amdgpu_video_codec_info; #define amdgpu_sriov_xgmi_ta_ext_peer_link_en(adev) \ ((adev)->virt.gim_feature & AMDGIM_FEATURE_XGMI_TA_EXT_PEER_LINK) +#define amdgpu_sriov_xgmi_connected_to_cpu(adev) \ +((adev)->virt.gim_feature & AMDGIM_FEATURE_XGMI_CONNECTED_TO_CPU) + static inline bool is_virtual_machine(void) { #if defined(CONFIG_X86) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h index a841f342a3eb..847cfd1fd004 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h @@ -161,7 +161,8 @@ union amd_sriov_msg_feature_flags { uint32_t ras_telemetry : 1; uint32_t ras_cper : 1; uint32_t xgmi_ta_ext_peer_link : 1; - uint32_t reserved : 19; + uint32_t xgmi_connected_to_cpu : 1; + uint32_t reserved : 18; } flags; uint32_t all; }; -- cgit v1.2.3 From 6d215051d272f749f5a31475d11ec1c9968bcca5 Mon Sep 17 00:00:00 2001 From: Likun Gao Date: Wed, 21 Jan 2026 14:56:54 +0800 Subject: drm/amdgpu: Add lsdma v7_1_0 ip headers Add header files for lsdma v7_1_0 register offsets and shift masks Signed-off-by: Likun Gao Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- .../include/asic_reg/lsdma/lsdma_7_1_0_offset.h | 44 +++++++++ .../include/asic_reg/lsdma/lsdma_7_1_0_sh_mask.h | 105 +++++++++++++++++++++ 2 files changed, 149 insertions(+) create mode 100644 drivers/gpu/drm/amd/include/asic_reg/lsdma/lsdma_7_1_0_offset.h create mode 100644 drivers/gpu/drm/amd/include/asic_reg/lsdma/lsdma_7_1_0_sh_mask.h diff --git a/drivers/gpu/drm/amd/include/asic_reg/lsdma/lsdma_7_1_0_offset.h b/drivers/gpu/drm/amd/include/asic_reg/lsdma/lsdma_7_1_0_offset.h new file mode 100644 index 000000000000..fa4e42a3ae9f --- /dev/null +++ b/drivers/gpu/drm/amd/include/asic_reg/lsdma/lsdma_7_1_0_offset.h @@ -0,0 +1,44 @@ +/* + * Copyright 2026 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef _lsdma_7_1_0_OFFSET_HEADER +#define _lsdma_7_1_0_OFFSET_HEADER + +#define regLSDMA_PIO_SRC_ADDR_LO 0x0080 +#define regLSDMA_PIO_SRC_ADDR_LO_BASE_IDX 0 +#define regLSDMA_PIO_SRC_ADDR_HI 0x0081 +#define regLSDMA_PIO_SRC_ADDR_HI_BASE_IDX 0 +#define regLSDMA_PIO_DST_ADDR_LO 0x0082 +#define regLSDMA_PIO_DST_ADDR_LO_BASE_IDX 0 +#define regLSDMA_PIO_DST_ADDR_HI 0x0083 +#define regLSDMA_PIO_DST_ADDR_HI_BASE_IDX 0 +#define regLSDMA_PIO_COMMAND 0x0084 +#define regLSDMA_PIO_COMMAND_BASE_IDX 0 +#define regLSDMA_PIO_CONSTFILL_DATA 0x0085 +#define regLSDMA_PIO_CONSTFILL_DATA_BASE_IDX 0 +#define regLSDMA_PIO_CONTROL 0x0086 +#define regLSDMA_PIO_CONTROL_BASE_IDX 0 + +#define regLSDMA_PIO_STATUS 0x008a +#define regLSDMA_PIO_STATUS_BASE_IDX 0 + +#endif diff --git a/drivers/gpu/drm/amd/include/asic_reg/lsdma/lsdma_7_1_0_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/lsdma/lsdma_7_1_0_sh_mask.h new file mode 100644 index 000000000000..cf83dacf4acf --- /dev/null +++ b/drivers/gpu/drm/amd/include/asic_reg/lsdma/lsdma_7_1_0_sh_mask.h @@ -0,0 +1,105 @@ +/* + * Copyright 2026 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef _lsdma_7_1_0_SH_MASK_HEADER +#define _lsdma_7_1_0_SH_MASK_HEADER + + +// addressBlock: lsdma0_lsdma0dec +//LSDMA_PIO_STATUS +#define LSDMA_PIO_STATUS__CMD_IN_FIFO__SHIFT 0x0 +#define LSDMA_PIO_STATUS__CMD_PROCESSING__SHIFT 0x3 +#define LSDMA_PIO_STATUS__ERROR_INVALID_ADDR__SHIFT 0xb +#define LSDMA_PIO_STATUS__ERROR_ZERO_COUNT__SHIFT 0xc +#define LSDMA_PIO_STATUS__ERROR_DRAM_ECC__SHIFT 0xd +#define LSDMA_PIO_STATUS__ERROR_SRAM_ECC__SHIFT 0xe +#define LSDMA_PIO_STATUS__ERROR_WRRET_NACK_GEN_ERR__SHIFT 0xf +#define LSDMA_PIO_STATUS__ERROR_RDRET_NACK_GEN_ERR__SHIFT 0x10 +#define LSDMA_PIO_STATUS__ERROR_WRRET_NACK_PRT__SHIFT 0x11 +#define LSDMA_PIO_STATUS__ERROR_RDRET_NACK_PRT__SHIFT 0x12 +#define LSDMA_PIO_STATUS__ERROR_REQ_DROP__SHIFT 0x13 +#define LSDMA_PIO_STATUS__PIO_FIFO_EMPTY__SHIFT 0x1c +#define LSDMA_PIO_STATUS__PIO_FIFO_FULL__SHIFT 0x1d +#define LSDMA_PIO_STATUS__PIO_IDLE__SHIFT 0x1f +#define LSDMA_PIO_STATUS__CMD_IN_FIFO_MASK 0x00000007L +#define LSDMA_PIO_STATUS__CMD_PROCESSING_MASK 0x000003F8L +#define LSDMA_PIO_STATUS__ERROR_INVALID_ADDR_MASK 0x00000800L +#define LSDMA_PIO_STATUS__ERROR_ZERO_COUNT_MASK 0x00001000L +#define LSDMA_PIO_STATUS__ERROR_DRAM_ECC_MASK 0x00002000L +#define LSDMA_PIO_STATUS__ERROR_SRAM_ECC_MASK 0x00004000L +#define LSDMA_PIO_STATUS__ERROR_WRRET_NACK_GEN_ERR_MASK 0x00008000L +#define LSDMA_PIO_STATUS__ERROR_RDRET_NACK_GEN_ERR_MASK 0x00010000L +#define LSDMA_PIO_STATUS__ERROR_WRRET_NACK_PRT_MASK 0x00020000L +#define LSDMA_PIO_STATUS__ERROR_RDRET_NACK_PRT_MASK 0x00040000L +#define LSDMA_PIO_STATUS__ERROR_REQ_DROP_MASK 0x00080000L +#define LSDMA_PIO_STATUS__PIO_FIFO_EMPTY_MASK 0x10000000L +#define LSDMA_PIO_STATUS__PIO_FIFO_FULL_MASK 0x20000000L +#define LSDMA_PIO_STATUS__PIO_IDLE_MASK 0x80000000L +//LSDMA_PIO_SRC_ADDR_LO +#define LSDMA_PIO_SRC_ADDR_LO__SRC_ADDR_LO__SHIFT 0x0 +#define LSDMA_PIO_SRC_ADDR_LO__SRC_ADDR_LO_MASK 0xFFFFFFFFL +//LSDMA_PIO_SRC_ADDR_HI +#define LSDMA_PIO_SRC_ADDR_HI__SRC_ADDR_HI__SHIFT 0x0 +#define LSDMA_PIO_SRC_ADDR_HI__SRC_ADDR_HI_MASK 0xFFFFFFFFL +//LSDMA_PIO_DST_ADDR_LO +#define LSDMA_PIO_DST_ADDR_LO__DST_ADDR_LO__SHIFT 0x0 +#define LSDMA_PIO_DST_ADDR_LO__DST_ADDR_LO_MASK 0xFFFFFFFFL +//LSDMA_PIO_DST_ADDR_HI +#define LSDMA_PIO_DST_ADDR_HI__DST_ADDR_HI__SHIFT 0x0 +#define LSDMA_PIO_DST_ADDR_HI__DST_ADDR_HI_MASK 0xFFFFFFFFL +//LSDMA_PIO_CONTROL +#define LSDMA_PIO_CONTROL__VMID__SHIFT 0x0 +#define LSDMA_PIO_CONTROL__DST_GPA__SHIFT 0x4 +#define LSDMA_PIO_CONTROL__DST_SYS__SHIFT 0x5 +#define LSDMA_PIO_CONTROL__DST_GCC__SHIFT 0x6 +#define LSDMA_PIO_CONTROL__DST_SNOOP__SHIFT 0x7 +#define LSDMA_PIO_CONTROL__DST_REUSE_HINT__SHIFT 0x8 +#define LSDMA_PIO_CONTROL__DST_COMP_EN__SHIFT 0xa +#define LSDMA_PIO_CONTROL__SRC_GPA__SHIFT 0x14 +#define LSDMA_PIO_CONTROL__SRC_SYS__SHIFT 0x15 +#define LSDMA_PIO_CONTROL__SRC_SNOOP__SHIFT 0x17 +#define LSDMA_PIO_CONTROL__SRC_REUSE_HINT__SHIFT 0x18 +#define LSDMA_PIO_CONTROL__SRC_COMP_EN__SHIFT 0x1a +#define LSDMA_PIO_CONTROL__VMID_MASK 0x0000000FL +#define LSDMA_PIO_CONTROL__DST_GPA_MASK 0x00000010L +#define LSDMA_PIO_CONTROL__DST_SYS_MASK 0x00000020L +#define LSDMA_PIO_CONTROL__DST_GCC_MASK 0x00000040L +#define LSDMA_PIO_CONTROL__DST_SNOOP_MASK 0x00000080L +#define LSDMA_PIO_CONTROL__DST_REUSE_HINT_MASK 0x00000300L +#define LSDMA_PIO_CONTROL__DST_COMP_EN_MASK 0x00000400L +#define LSDMA_PIO_CONTROL__SRC_GPA_MASK 0x00100000L +#define LSDMA_PIO_CONTROL__SRC_SYS_MASK 0x00200000L +#define LSDMA_PIO_CONTROL__SRC_SNOOP_MASK 0x00800000L +#define LSDMA_PIO_CONTROL__SRC_REUSE_HINT_MASK 0x03000000L +#define LSDMA_PIO_CONTROL__SRC_COMP_EN_MASK 0x04000000L +//LSDMA_PIO_COMMAND +#define LSDMA_PIO_COMMAND__COUNT__SHIFT 0x0 +#define LSDMA_PIO_COMMAND__RAW_WAIT__SHIFT 0x1e +#define LSDMA_PIO_COMMAND__CONSTANT_FILL__SHIFT 0x1f +#define LSDMA_PIO_COMMAND__COUNT_MASK 0x03FFFFFFL +#define LSDMA_PIO_COMMAND__RAW_WAIT_MASK 0x40000000L +#define LSDMA_PIO_COMMAND__CONSTANT_FILL_MASK 0x80000000L +//LSDMA_PIO_CONSTFILL_DATA +#define LSDMA_PIO_CONSTFILL_DATA__DATA__SHIFT 0x0 +#define LSDMA_PIO_CONSTFILL_DATA__DATA_MASK 0xFFFFFFFFL + +#endif -- cgit v1.2.3 From 0cdd5a0d68b452bf2e34c018b39ad454b979bd65 Mon Sep 17 00:00:00 2001 From: Likun Gao Date: Wed, 21 Jan 2026 15:03:25 +0800 Subject: drm/amdgpu: add support for lsdma v7_1 Add support for LSDMA v7_1_0. Signed-off-by: Likun Gao Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/Makefile | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 4 ++ drivers/gpu/drm/amd/amdgpu/lsdma_v7_1.c | 99 +++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/lsdma_v7_1.h | 31 +++++++++ 4 files changed, 135 insertions(+), 1 deletion(-) create mode 100644 drivers/gpu/drm/amd/amdgpu/lsdma_v7_1.c create mode 100644 drivers/gpu/drm/amd/amdgpu/lsdma_v7_1.h diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile index 006d49d6b4af..9c9c73b73ac8 100644 --- a/drivers/gpu/drm/amd/amdgpu/Makefile +++ b/drivers/gpu/drm/amd/amdgpu/Makefile @@ -86,7 +86,7 @@ amdgpu-y += \ nbio_v7_2.o hdp_v4_0.o hdp_v5_0.o aldebaran_reg_init.o aldebaran.o soc21.o soc24.o \ sienna_cichlid.o smu_v13_0_10.o nbio_v4_3.o hdp_v6_0.o nbio_v7_7.o hdp_v5_2.o lsdma_v6_0.o \ nbio_v7_9.o aqua_vanjaram.o nbio_v7_11.o lsdma_v7_0.o hdp_v7_0.o nbif_v6_3_1.o \ - cyan_skillfish_reg_init.o soc_v1_0.o + cyan_skillfish_reg_init.o soc_v1_0.o lsdma_v7_1.o # add DF block amdgpu-y += \ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index 9b0a3107bc90..2669b0346a60 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -87,6 +87,7 @@ #include "sdma_v7_1.h" #include "lsdma_v6_0.h" #include "lsdma_v7_0.h" +#include "lsdma_v7_1.h" #include "vcn_v2_0.h" #include "jpeg_v2_0.h" #include "vcn_v3_0.h" @@ -3308,6 +3309,9 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) case IP_VERSION(7, 0, 1): adev->lsdma.funcs = &lsdma_v7_0_funcs; break; + case IP_VERSION(7, 1, 0): + adev->lsdma.funcs = &lsdma_v7_1_funcs; + break; default: break; } diff --git a/drivers/gpu/drm/amd/amdgpu/lsdma_v7_1.c b/drivers/gpu/drm/amd/amdgpu/lsdma_v7_1.c new file mode 100644 index 000000000000..d93a0e65ce7d --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/lsdma_v7_1.c @@ -0,0 +1,99 @@ +/* + * Copyright 2026 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include "amdgpu.h" +#include "lsdma_v7_1.h" +#include "amdgpu_lsdma.h" + +#include "lsdma/lsdma_7_1_0_offset.h" +#include "lsdma/lsdma_7_1_0_sh_mask.h" + +static int lsdma_v7_1_wait_pio_status(struct amdgpu_device *adev) +{ + return amdgpu_lsdma_wait_for(adev, SOC15_REG_OFFSET(LSDMA, 0, regLSDMA_PIO_STATUS), + LSDMA_PIO_STATUS__PIO_IDLE_MASK | LSDMA_PIO_STATUS__PIO_FIFO_EMPTY_MASK, + LSDMA_PIO_STATUS__PIO_IDLE_MASK | LSDMA_PIO_STATUS__PIO_FIFO_EMPTY_MASK); +} + +static int lsdma_v7_1_copy_mem(struct amdgpu_device *adev, + uint64_t src_addr, + uint64_t dst_addr, + uint64_t size) +{ + int ret; + uint32_t tmp; + + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_SRC_ADDR_LO, lower_32_bits(src_addr)); + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_SRC_ADDR_HI, upper_32_bits(src_addr)); + + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_DST_ADDR_LO, lower_32_bits(dst_addr)); + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_DST_ADDR_HI, upper_32_bits(dst_addr)); + + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_CONTROL, 0x0); + + tmp = RREG32_SOC15(LSDMA, 0, regLSDMA_PIO_COMMAND); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, COUNT, size); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, RAW_WAIT, 0); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, CONSTANT_FILL, 0); + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_COMMAND, tmp); + + ret = lsdma_v7_1_wait_pio_status(adev); + if (ret) + dev_err(adev->dev, "LSDMA PIO failed to copy memory!\n"); + + return ret; +} + +static int lsdma_v7_1_fill_mem(struct amdgpu_device *adev, + uint64_t dst_addr, + uint32_t data, + uint64_t size) +{ + int ret; + uint32_t tmp; + + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_CONSTFILL_DATA, data); + + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_DST_ADDR_LO, lower_32_bits(dst_addr)); + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_DST_ADDR_HI, upper_32_bits(dst_addr)); + + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_CONTROL, 0x0); + + tmp = RREG32_SOC15(LSDMA, 0, regLSDMA_PIO_COMMAND); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, COUNT, size); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, RAW_WAIT, 0); + tmp = REG_SET_FIELD(tmp, LSDMA_PIO_COMMAND, CONSTANT_FILL, 1); + WREG32_SOC15(LSDMA, 0, regLSDMA_PIO_COMMAND, tmp); + + ret = lsdma_v7_1_wait_pio_status(adev); + if (ret) + dev_err(adev->dev, "LSDMA PIO failed to fill memory!\n"); + + return ret; +} + +const struct amdgpu_lsdma_funcs lsdma_v7_1_funcs = { + .copy_mem = lsdma_v7_1_copy_mem, + .fill_mem = lsdma_v7_1_fill_mem, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/lsdma_v7_1.h b/drivers/gpu/drm/amd/amdgpu/lsdma_v7_1.h new file mode 100644 index 000000000000..3d1ab605849a --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/lsdma_v7_1.h @@ -0,0 +1,31 @@ +/* + * Copyright 2026 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __LSDMA_V7_1_H__ +#define __LSDMA_V7_1_H__ + +#include "soc15_common.h" + +extern const struct amdgpu_lsdma_funcs lsdma_v7_1_funcs; + +#endif /* __LSDMA_V7_1_H__ */ -- cgit v1.2.3 From 50808826a64b4957b7088c789e539dd0a75a1560 Mon Sep 17 00:00:00 2001 From: Likun Gao Date: Fri, 26 Jul 2024 09:33:17 +0800 Subject: drm/amdgpu: fix DF NULL pointer issue for soc24 If DF function not initialized, NULL pointer issue will happen on soc24. Signed-off-by: Likun Gao Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/soc24.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/soc24.c b/drivers/gpu/drm/amd/amdgpu/soc24.c index 308f32daa780..d1adf19a51c4 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc24.c +++ b/drivers/gpu/drm/amd/amdgpu/soc24.c @@ -478,7 +478,7 @@ static int soc24_common_hw_init(struct amdgpu_ip_block *ip_block) if (adev->nbio.funcs->remap_hdp_registers) adev->nbio.funcs->remap_hdp_registers(adev); - if (adev->df.funcs->hw_init) + if (adev->df.funcs && adev->df.funcs->hw_init) adev->df.funcs->hw_init(adev); /* enable the doorbell aperture */ -- cgit v1.2.3 From 41c61e60f82d55e9a3a3f3f04d192d732230a91c Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Tue, 2 Dec 2025 21:36:19 +0800 Subject: drm/amdgpu: Handle IH v7_1 reg offset differences IH v7_1 changes the offsets of some registers relative to IH v7_0. Introduce IH v7_1-specific register access Signed-off-by: Hawking Zhang Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/ih_v7_0.c | 36 +++++++++++++++++++++++++++++------- 1 file changed, 29 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/ih_v7_0.c b/drivers/gpu/drm/amd/amdgpu/ih_v7_0.c index 451828bf583e..1fbe904f4223 100644 --- a/drivers/gpu/drm/amd/amdgpu/ih_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/ih_v7_0.c @@ -289,6 +289,13 @@ static uint32_t ih_v7_0_setup_retry_doorbell(u32 doorbell_index) return val; } +#define regIH_RING1_CLIENT_CFG_INDEX_V7_1 0x122 +#define regIH_RING1_CLIENT_CFG_INDEX_V7_1_BASE_IDX 0 +#define regIH_RING1_CLIENT_CFG_DATA_V7_1 0x123 +#define regIH_RING1_CLIENT_CFG_DATA_V7_1_BASE_IDX 0 +#define regIH_CHICKEN_V7_1 0x129 +#define regIH_CHICKEN_V7_1_BASE_IDX 0 + /** * ih_v7_0_irq_init - init and enable the interrupt ring * @@ -307,6 +314,7 @@ static int ih_v7_0_irq_init(struct amdgpu_device *adev) u32 tmp; int ret; int i; + u32 reg_addr; /* disable irqs */ ret = ih_v7_0_toggle_interrupts(adev, false); @@ -318,10 +326,15 @@ static int ih_v7_0_irq_init(struct amdgpu_device *adev) if (unlikely((adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) || (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO))) { if (ih[0]->use_bus_addr) { - ih_chicken = RREG32_SOC15(OSSSYS, 0, regIH_CHICKEN); + if (amdgpu_ip_version(adev, OSSSYS_HWIP, 0) == IP_VERSION(7, 1, 0)) + reg_addr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_CHICKEN_V7_1); + else + reg_addr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_CHICKEN); + ih_chicken = RREG32(reg_addr); + /* The reg fields definitions are identical in ih v7_0 and ih v7_1 */ ih_chicken = REG_SET_FIELD(ih_chicken, IH_CHICKEN, MC_SPACE_GPA_ENABLE, 1); - WREG32_SOC15(OSSSYS, 0, regIH_CHICKEN, ih_chicken); + WREG32(reg_addr, ih_chicken); } } @@ -358,17 +371,26 @@ static int ih_v7_0_irq_init(struct amdgpu_device *adev) /* Redirect the interrupts to IH RB1 for dGPU */ if (adev->irq.ih1.ring_size) { - tmp = RREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_INDEX); + if (amdgpu_ip_version(adev, OSSSYS_HWIP, 0) == IP_VERSION(7, 1, 0)) + reg_addr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RING1_CLIENT_CFG_INDEX_V7_1); + else + reg_addr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RING1_CLIENT_CFG_INDEX); + tmp = RREG32(reg_addr); + /* The reg fields definitions are identical in ih v7_0 and ih v7_1 */ tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_INDEX, INDEX, 0); - WREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_INDEX, tmp); + WREG32(reg_addr, tmp); - tmp = RREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_DATA); + if (amdgpu_ip_version(adev, OSSSYS_HWIP, 0) == IP_VERSION(7, 1, 0)) + reg_addr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RING1_CLIENT_CFG_DATA_V7_1); + else + reg_addr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RING1_CLIENT_CFG_DATA); + tmp = RREG32(reg_addr); + /* The reg fields definitions are identical in ih v7_0 and ih v7_1 */ tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA, CLIENT_ID, 0xa); tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA, SOURCE_ID, 0x0); tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA, SOURCE_ID_MATCH_ENABLE, 0x1); - - WREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_DATA, tmp); + WREG32(reg_addr, tmp); } pci_set_master(adev->pdev); -- cgit v1.2.3 From 6952ee37fad0fbccc98266ac61fce33ecd9d7338 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 18 Nov 2025 17:41:02 -0500 Subject: drm/amdgpu: use common defines for GMC 12.1 HUB faults Use proper definitions rather than a number. Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h | 4 ++++ drivers/gpu/drm/amd/amdgpu/gmc_v12_1.c | 2 +- 2 files changed, 5 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h index 0e8a52d96573..b9fdc3276e81 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h @@ -94,6 +94,10 @@ enum amdgpu_memory_partition { #define AMDGPU_GMC9_FAULT_SOURCE_DATA_WRITE 0x20 #define AMDGPU_GMC9_FAULT_SOURCE_DATA_EXE 0x10 +#define AMDGPU_GMC121_FAULT_SOURCE_DATA_READ 0x400000 +#define AMDGPU_GMC121_FAULT_SOURCE_DATA_WRITE 0x200000 +#define AMDGPU_GMC121_FAULT_SOURCE_DATA_EXE 0x100000 + /* * GMC page fault information */ diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v12_1.c b/drivers/gpu/drm/amd/amdgpu/gmc_v12_1.c index dc8865c5879c..c49112d8300e 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v12_1.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v12_1.c @@ -121,7 +121,7 @@ static int gmc_v12_1_process_interrupt(struct amdgpu_device *adev, if (entry->src_id == UTCL2_1_0__SRCID__RETRY) { retry_fault = true; - write_fault = !!(entry->src_data[1] & 0x200000); + write_fault = !!(entry->src_data[1] & AMDGPU_GMC121_FAULT_SOURCE_DATA_WRITE); } if (entry->client_id == SOC_V1_0_IH_CLIENTID_VMC) { -- cgit v1.2.3 From eb737fb303dd73ceb0f646fc2e696595e576056f Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Tue, 25 Nov 2025 01:42:42 +0800 Subject: drm/amdgpu: Add default reset method for soc_v1_0 Add mode2 as default reset method for soc_v1_0 v2: Remove unnecessary overrides while selecting reset method (Lijo) v4: Add dev_warn_once (Lijo) Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/soc_v1_0.c | 24 ++++++++++++++++++++++++ 1 file changed, 24 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c b/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c index 26e7566a5479..bd7043729e6a 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c @@ -229,8 +229,31 @@ static bool soc_v1_0_need_reset_on_init(struct amdgpu_device *adev) return false; } +static enum amd_reset_method +soc_v1_0_asic_reset_method(struct amdgpu_device *adev) +{ + if ((adev->gmc.xgmi.supported && adev->gmc.xgmi.connected_to_cpu) || + (amdgpu_ip_version(adev, MP1_HWIP, 0) == IP_VERSION(15, 0, 8))) { + if (amdgpu_reset_method != -1) + dev_warn_once(adev->dev, "Reset override isn't supported, using Mode2 instead.\n"); + + return AMD_RESET_METHOD_MODE2; + } + + return amdgpu_reset_method; +} + static int soc_v1_0_asic_reset(struct amdgpu_device *adev) { + switch (soc_v1_0_asic_reset_method(adev)) { + case AMD_RESET_METHOD_MODE2: + dev_info(adev->dev, "MODE2 reset\n"); + return amdgpu_dpm_mode2_reset(adev); + default: + dev_info(adev->dev, "Invalid reset method Not supported\n"); + return -EOPNOTSUPP; + } + return 0; } @@ -244,6 +267,7 @@ static const struct amdgpu_asic_funcs soc_v1_0_asic_funcs = { .need_reset_on_init = &soc_v1_0_need_reset_on_init, .encode_ext_smn_addressing = &soc_v1_0_encode_ext_smn_addressing, .reset = soc_v1_0_asic_reset, + .reset_method = &soc_v1_0_asic_reset_method, }; static int soc_v1_0_common_early_init(struct amdgpu_ip_block *ip_block) -- cgit v1.2.3 From d49f8adcbcaaeef1c4081d93fd9c35b018ab0652 Mon Sep 17 00:00:00 2001 From: Asad Kamal Date: Tue, 2 Dec 2025 10:54:59 +0800 Subject: drm/amd/amdgpu: Disable reset on init for soc_v1_0 Return false from soc_v1_0_reset_on_init as psp is loaded with ifwi and sol register will be non zero on first load itself Signed-off-by: Asad Kamal Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/soc_v1_0.c | 11 ----------- 1 file changed, 11 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c b/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c index bd7043729e6a..43209eae81bb 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c @@ -214,17 +214,6 @@ static bool soc_v1_0_need_full_reset(struct amdgpu_device *adev) static bool soc_v1_0_need_reset_on_init(struct amdgpu_device *adev) { - u32 sol_reg; - - if (adev->flags & AMD_IS_APU) - return false; - - /* Check sOS sign of life register to confirm sys driver and sOS - * are already been loaded. - */ - sol_reg = RREG32_SOC15(MP0, 0, regMPASP_SMN_C2PMSG_81); - if (sol_reg) - return true; return false; } -- cgit v1.2.3 From 3ebe58535898cb0a842390ecde79b5bc63a5b777 Mon Sep 17 00:00:00 2001 From: Le Ma Date: Tue, 2 Dec 2025 19:46:36 +0800 Subject: drm/amdgpu: update GFX CGCG/LS flags for gfx 12.1 Update GFX CGCG flags and fix num_xcc assignment Signed-off-by: Le Ma Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/soc_v1_0.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c b/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c index 43209eae81bb..a0e10dbf0733 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c @@ -281,7 +281,8 @@ static int soc_v1_0_common_early_init(struct amdgpu_ip_block *ip_block) switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { case IP_VERSION(12, 1, 0): - adev->cg_flags = 0; + adev->cg_flags = AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS; adev->pg_flags = 0; adev->external_rev_id = adev->rev_id + 0x50; break; -- cgit v1.2.3 From 68af620c9a370863e973fa27f94b635a90b8a045 Mon Sep 17 00:00:00 2001 From: Le Ma Date: Fri, 12 Dec 2025 18:43:47 +0800 Subject: drm/amdgpu: bypass IMU ucode loading for MP0 15.0.8 For MP0 15.0.8, IMU ucode is part of IFWI and ASP would load it by default. Signed-off-by: Le Ma Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index 27b67da9fdac..d39b695cd925 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -3096,6 +3096,13 @@ static int psp_load_non_psp_fw(struct psp_context *psp) */ continue; + /* IMU ucode is part of IFWI and MP0 15.0.8 would load it */ + if (amdgpu_ip_version(adev, MP0_HWIP, 0) == + IP_VERSION(15, 0, 8) && + (ucode->ucode_id == AMDGPU_UCODE_ID_IMU_I || + ucode->ucode_id == AMDGPU_UCODE_ID_IMU_D)) + continue; + psp_print_fw_hdr(psp, ucode); ret = psp_execute_ip_fw_load(psp, ucode); -- cgit v1.2.3 From 1394a4926f4bcff0dc6ac6deea5beeb2844297f0 Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Wed, 28 Jan 2026 18:53:40 +0800 Subject: drm/amdgpu: fix shift-out-of-bounds when updating umc active mask UMC node_inst_num can exceed 32, causing (1 << node_inst_num) to shift a 32-bit int out of bounds Signed-off-by: Hawking Zhang Reviewed-by: Likun Gao Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index 2669b0346a60..5f8f0c841666 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -793,7 +793,7 @@ static void amdgpu_discovery_read_from_harvest_table(struct amdgpu_device *adev, struct harvest_table *harvest_info; u16 offset; int i; - uint32_t umc_harvest_config = 0; + u64 umc_harvest_config = 0; if (amdgpu_discovery_get_table_info(adev, &info, HARVEST_INFO)) return; @@ -850,7 +850,7 @@ static void amdgpu_discovery_read_from_harvest_table(struct amdgpu_device *adev, } } - adev->umc.active_mask = ((1 << adev->umc.node_inst_num) - 1) & + adev->umc.active_mask = ((1ULL << adev->umc.node_inst_num) - 1ULL) & ~umc_harvest_config; } -- cgit v1.2.3 From c52b6c3709258760bd0950a75e02195811203da1 Mon Sep 17 00:00:00 2001 From: Likun Gao Date: Fri, 6 Feb 2026 09:09:47 +0800 Subject: drm/amdgpu: correct SDMA instance number for soc v1_0 Calculate sdma instance number according to xcc_mask and num_inst_per_xcc, and correct adev->sdma.sdma_mask according to totally sdma instance number. Signed-off-by: Likun Gao Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/soc_v1_0.c | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c b/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c index a0e10dbf0733..90a69ef41c4b 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c @@ -823,7 +823,7 @@ int soc_v1_0_init_soc_config(struct amdgpu_device *adev) { int ret, i; int xcc_inst_per_aid = 4; - uint16_t xcc_mask; + uint16_t xcc_mask, sdma_mask = 0; xcc_mask = adev->gfx.xcc_mask; adev->aid_mask = 0; @@ -833,10 +833,12 @@ int soc_v1_0_init_soc_config(struct amdgpu_device *adev) } adev->sdma.num_inst_per_xcc = 2; - adev->sdma.num_instances = - NUM_XCC(adev->gfx.xcc_mask) * adev->sdma.num_inst_per_xcc; - adev->sdma.sdma_mask = - GENMASK(adev->sdma.num_instances - 1, 0); + for_each_inst(i, adev->gfx.xcc_mask) + sdma_mask |= + GENMASK(adev->sdma.num_inst_per_xcc - 1, 0) << + (i * adev->sdma.num_inst_per_xcc); + adev->sdma.sdma_mask = sdma_mask; + adev->sdma.num_instances = NUM_XCC(adev->sdma.sdma_mask); ret = soc_v1_0_xcp_mgr_init(adev); if (ret) -- cgit v1.2.3 From 57f765a7ef090c31a60b3e9b75236bb652c41dba Mon Sep 17 00:00:00 2001 From: Likun Gao Date: Thu, 26 Feb 2026 13:11:27 +0800 Subject: drm/amdgpu: change sdma doorbell size for soc v1 Change SDMA doorbel size to 14 per SDMA engine. Signed-off-by: Likun Gao Reviewed-by: Feifei Xu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/soc_v1_0.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c b/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c index 90a69ef41c4b..0c7759b82fa6 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/soc_v1_0.c @@ -57,7 +57,7 @@ static void soc_v1_0_doorbell_index_init(struct amdgpu_device *adev) adev->doorbell_index.userqueue_end = AMDGPU_SOC_V1_0_DOORBELL_USERQUEUE_END; adev->doorbell_index.xcc_doorbell_range = AMDGPU_SOC_V1_0_DOORBELL_XCC_RANGE; - adev->doorbell_index.sdma_doorbell_range = 20; + adev->doorbell_index.sdma_doorbell_range = 14; for (i = 0; i < adev->sdma.num_instances; i++) adev->doorbell_index.sdma_engine[i] = AMDGPU_SOC_V1_0_DOORBELL_sDMA_ENGINE_START + -- cgit v1.2.3 From d3f8a9eba034d4c223d516a5f31a011b428a7c21 Mon Sep 17 00:00:00 2001 From: Likun Gao Date: Sat, 14 Feb 2026 04:22:56 +0800 Subject: drm/amdgpu: disable rlc fw info print Disable to print RLC v2_5 related firmware information by default. Signed-off-by: Likun Gao Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c index 1ab61e7b35db..82333aeb4453 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c @@ -291,22 +291,22 @@ void amdgpu_ucode_print_rlc_hdr(const struct common_firmware_header *hdr) break; case 5: /* rlc_hdr v2_5 */ - DRM_INFO("rlc_iram_ucode_size_bytes: %u\n", + DRM_DEBUG("rlc_iram_ucode_size_bytes: %u\n", le32_to_cpu(rlc_hdr_v2_5->v2_2.rlc_iram_ucode_size_bytes)); - DRM_INFO("rlc_iram_ucode_offset_bytes: %u\n", + DRM_DEBUG("rlc_iram_ucode_offset_bytes: %u\n", le32_to_cpu(rlc_hdr_v2_5->v2_2.rlc_iram_ucode_offset_bytes)); - DRM_INFO("rlc_dram_ucode_size_bytes: %u\n", + DRM_DEBUG("rlc_dram_ucode_size_bytes: %u\n", le32_to_cpu(rlc_hdr_v2_5->v2_2.rlc_dram_ucode_size_bytes)); - DRM_INFO("rlc_dram_ucode_offset_bytes: %u\n", + DRM_DEBUG("rlc_dram_ucode_offset_bytes: %u\n", le32_to_cpu(rlc_hdr_v2_5->v2_2.rlc_dram_ucode_offset_bytes)); /* rlc_hdr v2_5 */ - DRM_INFO("rlc_1_iram_ucode_size_bytes: %u\n", + DRM_DEBUG("rlc_1_iram_ucode_size_bytes: %u\n", le32_to_cpu(rlc_hdr_v2_5->rlc_1_iram_ucode_size_bytes)); - DRM_INFO("rlc_1_iram_ucode_offset_bytes: %u\n", + DRM_DEBUG("rlc_1_iram_ucode_offset_bytes: %u\n", le32_to_cpu(rlc_hdr_v2_5->rlc_1_iram_ucode_offset_bytes)); - DRM_INFO("rlc_1_dram_ucode_size_bytes: %u\n", + DRM_DEBUG("rlc_1_dram_ucode_size_bytes: %u\n", le32_to_cpu(rlc_hdr_v2_5->rlc_1_dram_ucode_size_bytes)); - DRM_INFO("rlc_1_dram_ucode_offset_bytes: %u\n", + DRM_DEBUG("rlc_1_dram_ucode_offset_bytes: %u\n", le32_to_cpu(rlc_hdr_v2_5->rlc_1_dram_ucode_offset_bytes)); break; default: -- cgit v1.2.3 From cbbf33396c5b4b54e28e9865a28090eca7a13173 Mon Sep 17 00:00:00 2001 From: Likun Gao Date: Fri, 27 Feb 2026 15:16:09 +0800 Subject: drm/amdgpu: fix sysfs ip base addr with 64bit Correct the base addr value shown on sysfs with ignore reg_base_64, since the base_addr value have been over write when discovery_init. Signed-off-by: Likun Gao Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 9 ++------- 1 file changed, 2 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index 5f8f0c841666..f9f785c5d8ac 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -1215,13 +1215,8 @@ static int amdgpu_discovery_sysfs_ips(struct amdgpu_device *adev, ip_hw_instance->num_instance); ip_hw_instance->num_base_addresses = ip->num_base_address; - for (kk = 0; kk < ip_hw_instance->num_base_addresses; kk++) { - if (reg_base_64) - ip_hw_instance->base_addr[kk] = - lower_32_bits(le64_to_cpu(ip->base_address_64[kk])) & 0x3FFFFFFF; - else - ip_hw_instance->base_addr[kk] = ip->base_address[kk]; - } + for (kk = 0; kk < ip_hw_instance->num_base_addresses; kk++) + ip_hw_instance->base_addr[kk] = ip->base_address[kk]; kobject_init(&ip_hw_instance->kobj, &ip_hw_instance_ktype); ip_hw_instance->kobj.kset = &ip_hw_id->hw_id_kset; -- cgit v1.2.3 From 418b4e397570f1b0ffb6d5c8aa0872bc33071d49 Mon Sep 17 00:00:00 2001 From: Le Ma Date: Mon, 2 Mar 2026 15:43:20 +0800 Subject: drm/amdgpu: fix array out of bounds accesses for mes sw_fini The mes.fw[] is per-pipe resource shared accross xcc inst. And enlarge hung_queue array to max inst_pipes. Signed-off-by: Le Ma Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h | 6 +++--- drivers/gpu/drm/amd/amdgpu/mes_v12_1.c | 4 +++- 2 files changed, 6 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h index bcf2a067dc41..f80e3aca9c78 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h @@ -159,9 +159,9 @@ struct amdgpu_mes { int hung_queue_db_array_size; int hung_queue_hqd_info_offset; - struct amdgpu_bo *hung_queue_db_array_gpu_obj[AMDGPU_MAX_MES_PIPES]; - uint64_t hung_queue_db_array_gpu_addr[AMDGPU_MAX_MES_PIPES]; - void *hung_queue_db_array_cpu_addr[AMDGPU_MAX_MES_PIPES]; + struct amdgpu_bo *hung_queue_db_array_gpu_obj[AMDGPU_MAX_MES_INST_PIPES]; + uint64_t hung_queue_db_array_gpu_addr[AMDGPU_MAX_MES_INST_PIPES]; + void *hung_queue_db_array_cpu_addr[AMDGPU_MAX_MES_INST_PIPES]; /* cooperative dispatch */ bool enable_coop_mode; diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c b/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c index 7b8c670d0a9e..d8e4b52bdfd5 100644 --- a/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c +++ b/drivers/gpu/drm/amd/amdgpu/mes_v12_1.c @@ -1611,7 +1611,6 @@ static int mes_v12_1_sw_fini(struct amdgpu_ip_block *ip_block) amdgpu_bo_free_kernel(&adev->mes.eop_gpu_obj[inst], &adev->mes.eop_gpu_addr[inst], NULL); - amdgpu_ucode_release(&adev->mes.fw[inst]); if (adev->enable_uni_mes || pipe == AMDGPU_MES_SCHED_PIPE) { amdgpu_bo_free_kernel(&adev->mes.ring[inst].mqd_obj, @@ -1622,6 +1621,9 @@ static int mes_v12_1_sw_fini(struct amdgpu_ip_block *ip_block) } } + for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) + amdgpu_ucode_release(&adev->mes.fw[pipe]); + for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) { if (!adev->enable_uni_mes) { amdgpu_bo_free_kernel(&adev->gfx.kiq[xcc_id].ring.mqd_obj, -- cgit v1.2.3 From da16822ce5c32b5aca848eaea521936d4410d48c Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Mon, 9 Mar 2026 15:17:00 +0530 Subject: drm/amd/pm: Fix xgmi max speed reporting Fix XGMI max bitrate/width reporting on SMUv13.0.12 SOCs. The data format got changed when moved to static table from dynamic metrics table. Fixes: 1bec2f270766 ("drm/amd/pm: Fetch SMUv13.0.12 xgmi max speed/width") Signed-off-by: Lijo Lazar Reviewed-by: Asad Kamal Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c index 8610f3da514f..54a86eb77cd5 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c @@ -269,8 +269,9 @@ static void smu_v13_0_12_init_xgmi_data(struct smu_context *smu, int ret; if (smu_table->tables[SMU_TABLE_SMU_METRICS].version >= 0x13) { - max_width = (uint8_t)static_metrics->MaxXgmiWidth; - max_speed = (uint16_t)static_metrics->MaxXgmiBitrate; + max_width = (uint8_t)SMUQ10_ROUND(static_metrics->MaxXgmiWidth); + max_speed = + (uint16_t)SMUQ10_ROUND(static_metrics->MaxXgmiBitrate); ret = 0; } else { MetricsTable_t *metrics = (MetricsTable_t *)smu_table->metrics_table; -- cgit v1.2.3 From fae5984296b981c8cc3acca35b701c1f332a6cd8 Mon Sep 17 00:00:00 2001 From: Mario Limonciello Date: Tue, 10 Mar 2026 11:58:22 -0500 Subject: drm/amd: Set num IP blocks to 0 if discovery fails If discovery has failed for any reason (such as no support for a block) then there is no need to unwind all the IP blocks in fini. In this condition there can actually be failures during the unwind too. Reset num_ip_blocks to zero during failure path and skip the unnecessary cleanup path. Suggested-by: Lijo Lazar Reviewed-by: Lijo Lazar Signed-off-by: Mario Limonciello Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 +++- drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 2 +- 2 files changed, 4 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 99e5d1e0bf3c..711b4502653a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -1995,8 +1995,10 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev) break; default: r = amdgpu_discovery_set_ip_blocks(adev); - if (r) + if (r) { + adev->num_ip_blocks = 0; return r; + } break; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index 77e2133de5cf..7f19554b9ad1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -83,7 +83,7 @@ void amdgpu_driver_unload_kms(struct drm_device *dev) { struct amdgpu_device *adev = drm_to_adev(dev); - if (adev == NULL) + if (adev == NULL || !adev->num_ip_blocks) return; amdgpu_unregister_gpu_instance(adev); -- cgit v1.2.3 From 7a9419ab42699fd3d4c857ef81ae097d8d8d5899 Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 20 Jan 2026 13:09:52 +0100 Subject: drm/amdgpu: revert to old status lock handling v4 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit It turned out that protecting the status of each bo_va with a spinlock was just hiding problems instead of solving them. Revert the whole approach, add a separate stats_lock and lockdep assertions that the correct reservation lock is held all over the place. This not only allows for better checks if a state transition is properly protected by a lock, but also switching back to using list macros to iterate over the state of lists protected by the dma_resv lock of the root PD. v2: re-add missing check v3: split into two patches v4: re-apply by fixing holding the VM lock at the right places. Signed-off-by: Christian König Reviewed-by: Alex Deucher Reviewed-by: Sunil Khatri Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 8 +- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 146 +++++++++++------------------- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 15 +-- drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c | 4 - 4 files changed, 68 insertions(+), 105 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c index 7cdff206e7f3..6d964a6ee349 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c @@ -1040,12 +1040,12 @@ amdgpu_userq_bo_validate(struct amdgpu_device *adev, struct drm_exec *exec, struct amdgpu_bo *bo; int ret; - spin_lock(&vm->status_lock); + spin_lock(&vm->invalidated_lock); while (!list_empty(&vm->invalidated)) { bo_va = list_first_entry(&vm->invalidated, struct amdgpu_bo_va, base.vm_status); - spin_unlock(&vm->status_lock); + spin_unlock(&vm->invalidated_lock); bo = bo_va->base.bo; ret = drm_exec_prepare_obj(exec, &bo->tbo.base, 2); @@ -1062,9 +1062,9 @@ amdgpu_userq_bo_validate(struct amdgpu_device *adev, struct drm_exec *exec, if (ret) return ret; - spin_lock(&vm->status_lock); + spin_lock(&vm->invalidated_lock); } - spin_unlock(&vm->status_lock); + spin_unlock(&vm->invalidated_lock); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index f31b2116a7de..d927f3aa544e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -153,12 +153,10 @@ static void amdgpu_vm_bo_evicted(struct amdgpu_vm_bo_base *vm_bo) vm_bo->moved = true; amdgpu_vm_assert_locked(vm); - spin_lock(&vm_bo->vm->status_lock); if (bo->tbo.type == ttm_bo_type_kernel) list_move(&vm_bo->vm_status, &vm->evicted); else list_move_tail(&vm_bo->vm_status, &vm->evicted); - spin_unlock(&vm_bo->vm->status_lock); } /** * amdgpu_vm_bo_moved - vm_bo is moved @@ -171,9 +169,7 @@ static void amdgpu_vm_bo_evicted(struct amdgpu_vm_bo_base *vm_bo) static void amdgpu_vm_bo_moved(struct amdgpu_vm_bo_base *vm_bo) { amdgpu_vm_assert_locked(vm_bo->vm); - spin_lock(&vm_bo->vm->status_lock); list_move(&vm_bo->vm_status, &vm_bo->vm->moved); - spin_unlock(&vm_bo->vm->status_lock); } /** @@ -187,9 +183,7 @@ static void amdgpu_vm_bo_moved(struct amdgpu_vm_bo_base *vm_bo) static void amdgpu_vm_bo_idle(struct amdgpu_vm_bo_base *vm_bo) { amdgpu_vm_assert_locked(vm_bo->vm); - spin_lock(&vm_bo->vm->status_lock); list_move(&vm_bo->vm_status, &vm_bo->vm->idle); - spin_unlock(&vm_bo->vm->status_lock); vm_bo->moved = false; } @@ -203,9 +197,9 @@ static void amdgpu_vm_bo_idle(struct amdgpu_vm_bo_base *vm_bo) */ static void amdgpu_vm_bo_invalidated(struct amdgpu_vm_bo_base *vm_bo) { - spin_lock(&vm_bo->vm->status_lock); + spin_lock(&vm_bo->vm->invalidated_lock); list_move(&vm_bo->vm_status, &vm_bo->vm->invalidated); - spin_unlock(&vm_bo->vm->status_lock); + spin_unlock(&vm_bo->vm->invalidated_lock); } /** @@ -218,10 +212,9 @@ static void amdgpu_vm_bo_invalidated(struct amdgpu_vm_bo_base *vm_bo) */ static void amdgpu_vm_bo_evicted_user(struct amdgpu_vm_bo_base *vm_bo) { + amdgpu_vm_assert_locked(vm_bo->vm); vm_bo->moved = true; - spin_lock(&vm_bo->vm->status_lock); list_move(&vm_bo->vm_status, &vm_bo->vm->evicted_user); - spin_unlock(&vm_bo->vm->status_lock); } /** @@ -235,13 +228,10 @@ static void amdgpu_vm_bo_evicted_user(struct amdgpu_vm_bo_base *vm_bo) static void amdgpu_vm_bo_relocated(struct amdgpu_vm_bo_base *vm_bo) { amdgpu_vm_assert_locked(vm_bo->vm); - if (vm_bo->bo->parent) { - spin_lock(&vm_bo->vm->status_lock); + if (vm_bo->bo->parent) list_move(&vm_bo->vm_status, &vm_bo->vm->relocated); - spin_unlock(&vm_bo->vm->status_lock); - } else { + else amdgpu_vm_bo_idle(vm_bo); - } } /** @@ -255,9 +245,7 @@ static void amdgpu_vm_bo_relocated(struct amdgpu_vm_bo_base *vm_bo) static void amdgpu_vm_bo_done(struct amdgpu_vm_bo_base *vm_bo) { amdgpu_vm_assert_locked(vm_bo->vm); - spin_lock(&vm_bo->vm->status_lock); list_move(&vm_bo->vm_status, &vm_bo->vm->done); - spin_unlock(&vm_bo->vm->status_lock); } /** @@ -271,13 +259,13 @@ static void amdgpu_vm_bo_reset_state_machine(struct amdgpu_vm *vm) { struct amdgpu_vm_bo_base *vm_bo, *tmp; - amdgpu_vm_assert_locked(vm); - - spin_lock(&vm->status_lock); + spin_lock(&vm->invalidated_lock); list_splice_init(&vm->done, &vm->invalidated); list_for_each_entry(vm_bo, &vm->invalidated, vm_status) vm_bo->moved = true; + spin_unlock(&vm->invalidated_lock); + amdgpu_vm_assert_locked(vm); list_for_each_entry_safe(vm_bo, tmp, &vm->idle, vm_status) { struct amdgpu_bo *bo = vm_bo->bo; @@ -287,14 +275,13 @@ static void amdgpu_vm_bo_reset_state_machine(struct amdgpu_vm *vm) else if (bo->parent) list_move(&vm_bo->vm_status, &vm_bo->vm->relocated); } - spin_unlock(&vm->status_lock); } /** * amdgpu_vm_update_shared - helper to update shared memory stat * @base: base structure for tracking BO usage in a VM * - * Takes the vm status_lock and updates the shared memory stat. If the basic + * Takes the vm stats_lock and updates the shared memory stat. If the basic * stat changed (e.g. buffer was moved) amdgpu_vm_update_stats need to be called * as well. */ @@ -307,7 +294,7 @@ static void amdgpu_vm_update_shared(struct amdgpu_vm_bo_base *base) bool shared; dma_resv_assert_held(bo->tbo.base.resv); - spin_lock(&vm->status_lock); + spin_lock(&vm->stats_lock); shared = drm_gem_object_is_shared_for_memory_stats(&bo->tbo.base); if (base->shared != shared) { base->shared = shared; @@ -319,7 +306,7 @@ static void amdgpu_vm_update_shared(struct amdgpu_vm_bo_base *base) vm->stats[bo_memtype].drm.private += size; } } - spin_unlock(&vm->status_lock); + spin_unlock(&vm->stats_lock); } /** @@ -344,11 +331,11 @@ void amdgpu_vm_bo_update_shared(struct amdgpu_bo *bo) * be bo->tbo.resource * @sign: if we should add (+1) or subtract (-1) from the stat * - * Caller need to have the vm status_lock held. Useful for when multiple update + * Caller need to have the vm stats_lock held. Useful for when multiple update * need to happen at the same time. */ static void amdgpu_vm_update_stats_locked(struct amdgpu_vm_bo_base *base, - struct ttm_resource *res, int sign) + struct ttm_resource *res, int sign) { struct amdgpu_vm *vm = base->vm; struct amdgpu_bo *bo = base->bo; @@ -372,7 +359,8 @@ static void amdgpu_vm_update_stats_locked(struct amdgpu_vm_bo_base *base, */ if (bo->flags & AMDGPU_GEM_CREATE_DISCARDABLE) vm->stats[res_memtype].drm.purgeable += size; - if (!(bo->preferred_domains & amdgpu_mem_type_to_domain(res_memtype))) + if (!(bo->preferred_domains & + amdgpu_mem_type_to_domain(res_memtype))) vm->stats[bo_memtype].evicted += size; } } @@ -391,9 +379,9 @@ void amdgpu_vm_update_stats(struct amdgpu_vm_bo_base *base, { struct amdgpu_vm *vm = base->vm; - spin_lock(&vm->status_lock); + spin_lock(&vm->stats_lock); amdgpu_vm_update_stats_locked(base, res, sign); - spin_unlock(&vm->status_lock); + spin_unlock(&vm->stats_lock); } /** @@ -419,10 +407,10 @@ void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base, base->next = bo->vm_bo; bo->vm_bo = base; - spin_lock(&vm->status_lock); + spin_lock(&vm->stats_lock); base->shared = drm_gem_object_is_shared_for_memory_stats(&bo->tbo.base); amdgpu_vm_update_stats_locked(base, bo->tbo.resource, +1); - spin_unlock(&vm->status_lock); + spin_unlock(&vm->stats_lock); if (!amdgpu_vm_is_bo_always_valid(vm, bo)) return; @@ -481,25 +469,25 @@ int amdgpu_vm_lock_done_list(struct amdgpu_vm *vm, struct drm_exec *exec, int ret; /* We can only trust prev->next while holding the lock */ - spin_lock(&vm->status_lock); + spin_lock(&vm->invalidated_lock); while (!list_is_head(prev->next, &vm->done)) { bo_va = list_entry(prev->next, typeof(*bo_va), base.vm_status); bo = bo_va->base.bo; if (bo) { amdgpu_bo_ref(bo); - spin_unlock(&vm->status_lock); + spin_unlock(&vm->invalidated_lock); ret = drm_exec_prepare_obj(exec, &bo->tbo.base, 1); amdgpu_bo_unref(&bo); if (unlikely(ret)) return ret; - spin_lock(&vm->status_lock); + spin_lock(&vm->invalidated_lock); } prev = prev->next; } - spin_unlock(&vm->status_lock); + spin_unlock(&vm->invalidated_lock); return 0; } @@ -595,7 +583,7 @@ int amdgpu_vm_validate(struct amdgpu_device *adev, struct amdgpu_vm *vm, void *param) { uint64_t new_vm_generation = amdgpu_vm_generation(adev, vm); - struct amdgpu_vm_bo_base *bo_base; + struct amdgpu_vm_bo_base *bo_base, *tmp; struct amdgpu_bo *bo; int r; @@ -608,13 +596,7 @@ int amdgpu_vm_validate(struct amdgpu_device *adev, struct amdgpu_vm *vm, return r; } - spin_lock(&vm->status_lock); - while (!list_empty(&vm->evicted)) { - bo_base = list_first_entry(&vm->evicted, - struct amdgpu_vm_bo_base, - vm_status); - spin_unlock(&vm->status_lock); - + list_for_each_entry_safe(bo_base, tmp, &vm->evicted, vm_status) { bo = bo_base->bo; r = validate(param, bo); @@ -627,26 +609,21 @@ int amdgpu_vm_validate(struct amdgpu_device *adev, struct amdgpu_vm *vm, vm->update_funcs->map_table(to_amdgpu_bo_vm(bo)); amdgpu_vm_bo_relocated(bo_base); } - spin_lock(&vm->status_lock); } - while (ticket && !list_empty(&vm->evicted_user)) { - bo_base = list_first_entry(&vm->evicted_user, - struct amdgpu_vm_bo_base, - vm_status); - spin_unlock(&vm->status_lock); - bo = bo_base->bo; - dma_resv_assert_held(bo->tbo.base.resv); + if (ticket) { + list_for_each_entry_safe(bo_base, tmp, &vm->evicted_user, + vm_status) { + bo = bo_base->bo; + dma_resv_assert_held(bo->tbo.base.resv); - r = validate(param, bo); - if (r) - return r; - - amdgpu_vm_bo_invalidated(bo_base); + r = validate(param, bo); + if (r) + return r; - spin_lock(&vm->status_lock); + amdgpu_vm_bo_invalidated(bo_base); + } } - spin_unlock(&vm->status_lock); amdgpu_vm_eviction_lock(vm); vm->evicting = false; @@ -675,9 +652,7 @@ bool amdgpu_vm_ready(struct amdgpu_vm *vm) ret = !vm->evicting; amdgpu_vm_eviction_unlock(vm); - spin_lock(&vm->status_lock); ret &= list_empty(&vm->evicted); - spin_unlock(&vm->status_lock); spin_lock(&vm->immediate.lock); ret &= !vm->immediate.stopped; @@ -971,18 +946,13 @@ int amdgpu_vm_update_pdes(struct amdgpu_device *adev, struct amdgpu_vm *vm, bool immediate) { struct amdgpu_vm_update_params params; - struct amdgpu_vm_bo_base *entry; + struct amdgpu_vm_bo_base *entry, *tmp; bool flush_tlb_needed = false; - LIST_HEAD(relocated); int r, idx; amdgpu_vm_assert_locked(vm); - spin_lock(&vm->status_lock); - list_splice_init(&vm->relocated, &relocated); - spin_unlock(&vm->status_lock); - - if (list_empty(&relocated)) + if (list_empty(&vm->relocated)) return 0; if (!drm_dev_enter(adev_to_drm(adev), &idx)) @@ -998,7 +968,7 @@ int amdgpu_vm_update_pdes(struct amdgpu_device *adev, if (r) goto error; - list_for_each_entry(entry, &relocated, vm_status) { + list_for_each_entry(entry, &vm->relocated, vm_status) { /* vm_flush_needed after updating moved PDEs */ flush_tlb_needed |= entry->moved; @@ -1014,9 +984,7 @@ int amdgpu_vm_update_pdes(struct amdgpu_device *adev, if (flush_tlb_needed) atomic64_inc(&vm->tlb_seq); - while (!list_empty(&relocated)) { - entry = list_first_entry(&relocated, struct amdgpu_vm_bo_base, - vm_status); + list_for_each_entry_safe(entry, tmp, &vm->relocated, vm_status) { amdgpu_vm_bo_idle(entry); } @@ -1243,9 +1211,9 @@ error_free: void amdgpu_vm_get_memory(struct amdgpu_vm *vm, struct amdgpu_mem_stats stats[__AMDGPU_PL_NUM]) { - spin_lock(&vm->status_lock); + spin_lock(&vm->stats_lock); memcpy(stats, vm->stats, sizeof(*stats) * __AMDGPU_PL_NUM); - spin_unlock(&vm->status_lock); + spin_unlock(&vm->stats_lock); } /** @@ -1612,29 +1580,24 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev, struct amdgpu_vm *vm, struct ww_acquire_ctx *ticket) { - struct amdgpu_bo_va *bo_va; + struct amdgpu_bo_va *bo_va, *tmp; struct dma_resv *resv; bool clear, unlock; int r; - spin_lock(&vm->status_lock); - while (!list_empty(&vm->moved)) { - bo_va = list_first_entry(&vm->moved, struct amdgpu_bo_va, - base.vm_status); - spin_unlock(&vm->status_lock); - + list_for_each_entry_safe(bo_va, tmp, &vm->moved, base.vm_status) { /* Per VM BOs never need to bo cleared in the page tables */ r = amdgpu_vm_bo_update(adev, bo_va, false); if (r) return r; - spin_lock(&vm->status_lock); } + spin_lock(&vm->invalidated_lock); while (!list_empty(&vm->invalidated)) { bo_va = list_first_entry(&vm->invalidated, struct amdgpu_bo_va, base.vm_status); resv = bo_va->base.bo->tbo.base.resv; - spin_unlock(&vm->status_lock); + spin_unlock(&vm->invalidated_lock); /* Try to reserve the BO to avoid clearing its ptes */ if (!adev->debug_vm && dma_resv_trylock(resv)) { @@ -1666,9 +1629,9 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev, bo_va->base.bo->tbo.resource->mem_type == TTM_PL_SYSTEM)) amdgpu_vm_bo_evicted_user(&bo_va->base); - spin_lock(&vm->status_lock); + spin_lock(&vm->invalidated_lock); } - spin_unlock(&vm->status_lock); + spin_unlock(&vm->invalidated_lock); return 0; } @@ -2211,9 +2174,9 @@ void amdgpu_vm_bo_del(struct amdgpu_device *adev, } } - spin_lock(&vm->status_lock); + spin_lock(&vm->invalidated_lock); list_del(&bo_va->base.vm_status); - spin_unlock(&vm->status_lock); + spin_unlock(&vm->invalidated_lock); list_for_each_entry_safe(mapping, next, &bo_va->valids, list) { list_del(&mapping->list); @@ -2321,10 +2284,10 @@ void amdgpu_vm_bo_move(struct amdgpu_bo *bo, struct ttm_resource *new_mem, for (bo_base = bo->vm_bo; bo_base; bo_base = bo_base->next) { struct amdgpu_vm *vm = bo_base->vm; - spin_lock(&vm->status_lock); + spin_lock(&vm->stats_lock); amdgpu_vm_update_stats_locked(bo_base, bo->tbo.resource, -1); amdgpu_vm_update_stats_locked(bo_base, new_mem, +1); - spin_unlock(&vm->status_lock); + spin_unlock(&vm->stats_lock); } amdgpu_vm_bo_invalidate(bo, evicted); @@ -2593,11 +2556,12 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, INIT_LIST_HEAD(&vm->relocated); INIT_LIST_HEAD(&vm->moved); INIT_LIST_HEAD(&vm->idle); + spin_lock_init(&vm->invalidated_lock); INIT_LIST_HEAD(&vm->invalidated); - spin_lock_init(&vm->status_lock); INIT_LIST_HEAD(&vm->freed); INIT_LIST_HEAD(&vm->done); INIT_KFIFO(vm->faults); + spin_lock_init(&vm->stats_lock); r = amdgpu_vm_init_entities(adev, vm); if (r) @@ -3065,7 +3029,6 @@ void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m) amdgpu_vm_assert_locked(vm); - spin_lock(&vm->status_lock); seq_puts(m, "\tIdle BOs:\n"); list_for_each_entry_safe(bo_va, tmp, &vm->idle, base.vm_status) { if (!bo_va->base.bo) @@ -3103,11 +3066,13 @@ void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m) id = 0; seq_puts(m, "\tInvalidated BOs:\n"); + spin_lock(&vm->invalidated_lock); list_for_each_entry_safe(bo_va, tmp, &vm->invalidated, base.vm_status) { if (!bo_va->base.bo) continue; total_invalidated += amdgpu_bo_print_info(id++, bo_va->base.bo, m); } + spin_unlock(&vm->invalidated_lock); total_invalidated_objs = id; id = 0; @@ -3117,7 +3082,6 @@ void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m) continue; total_done += amdgpu_bo_print_info(id++, bo_va->base.bo, m); } - spin_unlock(&vm->status_lock); total_done_objs = id; seq_printf(m, "\tTotal idle size: %12lld\tobjs:\t%d\n", total_idle, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index dc4b0ec672ec..ae9449d5b00c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -205,11 +205,11 @@ struct amdgpu_vm_bo_base { /* protected by bo being reserved */ struct amdgpu_vm_bo_base *next; - /* protected by vm status_lock */ + /* protected by vm reservation and invalidated_lock */ struct list_head vm_status; /* if the bo is counted as shared in mem stats - * protected by vm status_lock */ + * protected by vm BO being reserved */ bool shared; /* protected by the BO being reserved */ @@ -345,10 +345,8 @@ struct amdgpu_vm { bool evicting; unsigned int saved_flags; - /* Lock to protect vm_bo add/del/move on all lists of vm */ - spinlock_t status_lock; - - /* Memory statistics for this vm, protected by status_lock */ + /* Memory statistics for this vm, protected by stats_lock */ + spinlock_t stats_lock; struct amdgpu_mem_stats stats[__AMDGPU_PL_NUM]; /* @@ -356,6 +354,8 @@ struct amdgpu_vm { * PDs, PTs or per VM BOs. The state transits are: * * evicted -> relocated (PDs, PTs) or moved (per VM BOs) -> idle + * + * Lists are protected by the root PD dma_resv lock. */ /* Per-VM and PT BOs who needs a validation */ @@ -376,7 +376,10 @@ struct amdgpu_vm { * state transits are: * * evicted_user or invalidated -> done + * + * Lists are protected by the invalidated_lock. */ + spinlock_t invalidated_lock; /* BOs for user mode queues that need a validation */ struct list_head evicted_user; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c index 31a437ce9570..7bdd664f0770 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c @@ -544,9 +544,7 @@ static void amdgpu_vm_pt_free(struct amdgpu_vm_bo_base *entry) entry->bo->vm_bo = NULL; ttm_bo_set_bulk_move(&entry->bo->tbo, NULL); - spin_lock(&entry->vm->status_lock); list_del(&entry->vm_status); - spin_unlock(&entry->vm->status_lock); amdgpu_bo_unref(&entry->bo); } @@ -590,7 +588,6 @@ static void amdgpu_vm_pt_add_list(struct amdgpu_vm_update_params *params, struct amdgpu_vm_pt_cursor seek; struct amdgpu_vm_bo_base *entry; - spin_lock(¶ms->vm->status_lock); for_each_amdgpu_vm_pt_dfs_safe(params->adev, params->vm, cursor, seek, entry) { if (entry && entry->bo) list_move(&entry->vm_status, ¶ms->tlb_flush_waitlist); @@ -598,7 +595,6 @@ static void amdgpu_vm_pt_add_list(struct amdgpu_vm_update_params *params, /* enter start node now */ list_move(&cursor->entry->vm_status, ¶ms->tlb_flush_waitlist); - spin_unlock(¶ms->vm->status_lock); } /** -- cgit v1.2.3